CN101389521A - Driving behavior estimation device, driving support device, vehicle evaluation system, driver model generation device, and driving behavior determination device - Google Patents

Abstract

An object of the present invention is to create a higher-precision driver model as an evaluation standard of a driving state in a normal state, and to estimate driving behavior. Collect driving state data (accelerator, brake, steering operation amount, vehicle speed, inter-vehicle distance, acceleration, etc. own vehicle information) while the driver is driving, extract the part of the driving state data that the driver is driving in a normal state, and automatically create Normal driver model. In addition, by using GMM (Gaussian Mixture Model) as the driver model, it is possible to easily generate a driver model for each driver, and further estimate and output the driving Operational action.

Description

Drive behavior estimating device, drive supporting device, vehicle evaluating system, driver's model generating apparatus, and drive behavior decision maker

Technical field

The present invention relates to drive behavior estimating device, drive supporting device, reach vehicle evaluating system, for example, use driver's model to drive behavior, infer, drive supporting, and the device of vehicle evaluation, system and as the generating apparatus of driver's model of the metewand of driving condition and used the evaluation of driving condition of driver's model and the device of drive supporting.

Background technology

About the modelling of vehicle driver's (driver) driver behavior and use and proposed various suggestions.

For example, in the technology of patent documentation 1 record, propose to estimate the technology of the degree of risk of point of crossing road by having used driver's model of fuzzy rule or neural network.

Patent documentation 1: the spy opens 2002-140786

In patent documentation 1 described technology, use fuzzy rule or neural network to constitute driver's model, therefore, need the study of the generation of fuzzy rule or back-propagating etc., can not easily generate driver's model.

In addition, in the prior art, can generate with common driver is the model of object, but is difficult to more correctly to show the feature of each driver's driver behavior.

Therefore, can not generate each driver's model of feature of a plurality of drivers of expression.

And then driver's model in the past is the model that is used to estimate the degree of risk of point of crossing road, therefore, is not driver's model of inferring the drive behavior of driver behavior.

Summary of the invention

First purpose of the present invention is and can easily generates, and uses driver's model of the driving feature that can more correctly show the driver, infers drive behavior.

In addition,, will be estimated as normal driving condition, itself and current driving condition will be compared, and can estimate current driving thus by the driving conditions such as driver behavior that driver's model is inferred by using patent documentation 1 described driver's model.

But, may not a normal driving condition that be decided to be for described driver by the driving condition that driver's model is inferred.

In addition, though specific driver steering vehicle data of collecting driving condition in fact, and generate driver's model in advance based on this, also may not be to have carried out the driving under the normal condition.

Therefore, second purpose of the present invention is to generate the more high-precision driver's model as the metewand of driving condition.

In addition, the 3rd purpose is to provide the evaluation of the more high-precision driving condition of carrying out that has used this driver's model and the drive supporting device of drive supporting.

In addition, the 4th purpose is driver's drive behavior is carried out the evaluation of more high-precision driving condition.

(1) in the described drive behavior estimating device of first aspect, this drive behavior estimating device is realized described first purpose as described below, possess: driver's model, it travels the time series data of detected N kind characteristic quantity as learning data with escort vehicle, and the summary that each data in the N dimension space exist distributed describes; Characteristic quantity obtains mechanism, and it obtains at least a above characteristic quantity except specific characteristic quantity x in the described N kind; Maximum posterior probability is calculated mechanism, and it is calculated with respect to the maximum posterior probability in described driver's model of the characteristic quantity of described acquisition; Output mechanism, it is based on described maximum posterior probability of calculating, and output is with respect to the presumed value of the described specific characteristic quantity x of the characteristic quantity of described acquisition.

(2) the described invention of second aspect is characterized in that in the described drive behavior estimating device of first aspect, and described N kind characteristic quantity comprises: for the n kind (time variation amount of the characteristic quantity of n＜N).

(3) the described invention of the third aspect first or the described drive behavior estimating device of second aspect in, it is characterized in that described characteristic quantity x comprises: the operational ton of the handling device of driver's direct control and the time variation amount of this operational ton.

(4) the described invention of fourth aspect is in the described drive behavior estimating device of first～third aspect, it is characterized in that, described driver's model as learning data, is described probability distribution that each data exist with the GMM (gauss hybrid models) that utilizes the EM algorithm to calculate with the time series data of described N kind characteristic quantity.

(5) aspect the 5th in the described invention, a kind of drive supporting device, it is characterized in that, possess: each described drive behavior estimating device in first～fourth aspect, its use: as characteristic quantity, used accelerator operation amount, brake service amount, from the speed of a motor vehicle of vehicle, with the accelerator of the vehicle headway of front vehicles with driver's model and drg driver's model, as described characteristic quantity x, infer accelerator operation amount and brake service amount; Running data obtains mechanism, and it obtains the speed of a motor vehicle and vehicle headway from vehicle; Ride control mechanism, it is for the running data of described acquisition, according to the accelerator operation amount of being inferred by described drive behavior estimating device, and brake service amount, control engine throttle and brake pedal carry out the automatic follow running to described front vehicles thus.

(6) aspect the 6th in the described invention, a kind of vehicle evaluating system, it is characterized in that, possess: each described drive behavior estimating device in first～fourth aspect, its use: as characteristic quantity, used accelerator operation amount, brake service amount, from the speed of a motor vehicle of vehicle, with the accelerator of the vehicle headway of front vehicles with driver's model and drg driver's model, as described characteristic quantity x, infer accelerator operation amount and brake service amount; Acquisition is as the mechanism of the vehicle performance data of the vehicle of evaluation object; Obtain the mechanism of simulation with running data and track model; The vehicle power computing mechanism, it is to being applicable to that by running data and track model with described acquisition described drive behavior estimating device obtains accelerator operation amount and brake service amount, inferring the trace that comprises as the acceleration/accel of the vehicle of described evaluation object; Assessing mechanism, it is by the trace of described estimated vehicle, estimates the rideability of described vehicle as evaluation object.

(7) in the described invention, realize described second purpose as described below aspect the 7th, a kind of driver's model generating apparatus is characterized in that possessing: the state decision mechanism, and it judges driver's state; Driver behavior information acquisition mechanism, it obtains the driver behavior information in the vehicle '; Driver's model generates mechanism, and it is based on the driver behavior information of described acquisition, generates the driver's model with driver's the cooresponding driver behavior of state.

(8) in the described invention of eight aspect, in described driver's model generating apparatus, it is characterized in that aspect the 7th described state decision mechanism judges at least whether driver's state is normal.

(9) aspect the 9th in the described invention, the 7th or the described driver's model of eight aspect generating apparatus in, it is characterized in that running data obtains mechanism, it detects specific running environment; Each stores described driver behavior information by running environment, and described driver's model generates mechanism, and each generates driver's model by described running environment.

(10) aspect the 90 in the described invention, in aspect the 7th～the 9th in each described driver's model generating apparatus, it is characterized in that, possess: Biont information obtains mechanism, it obtains driver's Biont information, and described state decision mechanism is judged driver's state based on the Biont information of described acquisition.

(11) in the described invention of the tenth one side, realize described the 3rd purpose as described above, a kind of drive supporting device is characterized in that, possesses: driver's model obtains mechanism, and it obtains driver's model of the driver behavior under normal condition; The driver behavior estimating mechanism, it uses driver's model of described acquisition, infers the driver behavior of driving usually with normal condition; Drive behavior decision mechanism, it judges driver's drive behavior by described driver behavior of inferring with based on the driver behavior of current driver behavior information; Drive supporting mechanism, it carries out and the cooresponding drive supporting of the drive behavior of described judgement.

(12) aspect the 12 in the described invention, in the described drive supporting device of the tenth one side, it is characterized in that, described driver's model obtains the driver model of mechanism by the driver behavior under the normal condition of pressing each generation of driving environment, obtains and the current cooresponding driver's model of running environment.

(13) aspect the 13 in the described invention, aspect the 11 or the 12 in the described drive supporting device, it is characterized in that, possess: driver condition decision mechanism, it judges driver's state by driver's Biont information, and described drive supporting mechanism carries out and the drive behavior of described judgement and the cooresponding drive supporting of driver condition of described judgement.

(14) aspect the 14 in the described invention, aspect the 11～the 13 in the described drive supporting device, it is characterized in that, described drive supporting mechanism is according to judging content, utilizes attention prompting, the information of voice or image to provide, at least a above drive supporting in the guiding in vibration, Rest area.

(15) in the described invention, realize described the 4th invention as described below aspect the 15, a kind of drive behavior decision maker is characterized in that possessing: driver's model obtains mechanism, and it obtains driver's model of the driver behavior under the normal condition; The driver behavior estimating mechanism, it uses driver's model of described acquisition, infers the driver behavior of driving usually with normal condition; Drive behavior decision mechanism, it judges driver's drive behavior by described driver behavior of inferring with based on the driver behavior of current driver behavior information.

In the invention aspect first aspect～6th, use is travelled the time series data of detected N kind characteristic quantity as learning data with escort vehicle, and driver's model that the summary that each data in the N dimension space exist is distributed and describes, calculate with respect to the maximum posterior probability in described driver's model of described characteristic quantity except specific characteristic quantity x, presumed value output as described specific characteristic quantity x, therefore, can easily generate, can stipulate more drive behavior near driver's driving feature.

Aspect the 7th～the described the present invention in the tenth aspect in, generate the driver's model with driver's the cooresponding driver behavior of state, therefore, can access more high-precision driver's model.

Aspect the 11～the 14 among described the present invention, usually the driver behavior of driving by the normal condition of inferring and based on the driver behavior of current driver behavior information with the driver's model that uses the driver behavior under the normal condition, judge driver's drive behavior, carry out and the cooresponding drive supporting of judging of drive behavior, therefore, can carry out the evaluation and the drive supporting of more high-precision driving condition.

Aspect the 15 in the described invention, usually the driver behavior of driving by the normal condition of inferring and based on the driver behavior of current driver behavior information with the driver's model that uses the driver behavior under the normal condition, judge driver's drive behavior, therefore, can estimate more high-precision driving condition.

Description of drawings

Fig. 1 be expression with based on the generation of driver's model of the drive behavior estimating device in the present embodiment with based on the figure that infers relevant principle of the drive behavior of driver's model of generation.

Fig. 2 is expression and the instruction diagram of inferring relevant principle based on the drive behavior of maximum posterior probability.

Fig. 3 is the instruction diagram of the structure of expression drive behavior estimating device.

Fig. 4 is the instruction diagram of expression with the running data of running data acquisition portion acquisition.

Fig. 5 is expression generates processing based on driver's model of driver's model generating unit a diagram of circuit.

Fig. 6 is that expression uses the driver's model that generates to infer the diagram of circuit of the processing of specific drive behavior.

Fig. 7 is the constructional drawing that is suitable for the drive supporting device of drive behavior estimating device.

Fig. 8 is expression ACC handles action with the automatic generation of driver's model a diagram of circuit.

Fig. 9 is the diagram of circuit of the action of expression ACC processing.

Figure 10 is the conceptual illustration figure of the summary of expression vehicle evaluating system.

Figure 11 is the constructional drawing of drive behavior estimating device.

Figure 12 is the instruction diagram of the summary of each data in the expression drive behavior estimating device.

Figure 13 is the diagram of circuit that the action of handling is estimated in the expression design.

Figure 14 is expression study with the trace of driving earlier and drives data, and estimates with the instruction diagram of the trace of driving earlier.

Figure 15 is the instruction diagram that the analog result of vehicle evaluating system has been used in expression.

Figure 16 is the constructional drawing that is suitable for the drive supporting device of the driver's model generating apparatus in second embodiment of the present application.

Figure 17 has been an illustration the instruction diagram that in information of vehicles acquisition portion, obtains from information of vehicles.

The instruction diagram of the vehicle-surroundings environmental information that obtains in Figure 18 has been the illustration vehicle-surroundings information acquisition portion.

The instruction diagram of the vehicle-surroundings environmental information that obtains in Figure 19 has been the illustration road information acquisition portion.

The instruction diagram of the vehicle-surroundings environmental information that Figure 20 is an illustration obtains in the Network Dept..

The instruction diagram of the Biont information that obtains in Figure 21 has been the illustration Biont information acquisition portion.

The instruction diagram of the information that Figure 22 is an illustration is provided by the information portion of providing, auxiliary content.

Figure 23 is the instruction diagram that concept nature is represented the memory contents of driver's model storage part.

Figure 24 is the instruction diagram that concept nature is represented the content of status data.

Figure 25 is the diagram of circuit that driver's model of driver's model of expression " common drive behavior " (just often) that generate the driver generates the processing action of handling.

Figure 26 is expression monitors the state that spirit changes by the change of heart rate a instruction diagram.

Figure 27 is the state that monitors that spirit changes is resolved in expression by Electrocardiographic long-range navigation thatch topology a instruction diagram.

Figure 28 is normal condition whether situation is judged in expression by the Biont information that obtains a instruction diagram.

Figure 29 is an illustration when the point of crossing right-hand corner from information of vehicles acquisition portion and from vehicle-surroundings environmental information acquisition portion obtain from information of vehicles with from the instruction diagram of vehicle-surroundings environmental information.

Figure 30 is the diagram of circuit that expression driver drive behavior monitors the processing action of handling.

Figure 31 is the instruction diagram by the detection of the setting of the status data of carrying out from information of vehicles with from the vehicle-surroundings environmental information and the situation that is fit to.

Figure 32 is the relatively instruction diagram of the operational ton (from information of vehicles) of the presumed value of the driver behavior amount (common driving) under the normal condition of driver's model efferent output and current driving of concept nature.

Figure 33 is the diagram of circuit of handling about the supervision of the Biont information of the driver in travelling.

Figure 34 is that the state by driver's eyes is judged the instruction diagram under the situation of normal condition and sleepiness, fatigue state.

Figure 35 is the diagram of circuit of the action handled of the drive supporting of expression driver's driving condition and Biont information situation.

The instruction diagram of the content of Figure 36 driver's that to be expression infer by drive behavior that obtains and Biont information situation state and the drive supporting that carries out corresponding to the driver's who infers state.

Among the figure: 10-driver model generating unit; 101-driver information acquisition portion; 102-running data acquisition portion; 103-simultaneously probability density distribution calculating parts; 104-simultaneously probability density function parameter storage parts; 11-drive behavior estimating portion; 111-driver information acquisition portion; 112-running data acquisition portion; 113-driver Model Selection portion; 114-maximum posterior probability calculating part; The presumed value efferent of 115-characteristic quantity X; 50-ECU; 51-from information of vehicles acquisition portion; 52-from vehicle-surroundings environmental information acquisition portion; 521-vehicle-surroundings information acquisition portion; 522-road information acquisition portion; 523-Network Dept.; 53-Biont information acquisition portion; 54-information provides portion; 55-driver model handling part; 551-driver model generating unit; 552-driver model storage part; 553-driver model efferent; 56-data store.

The specific embodiment

Below, with reference to Fig. 1～Figure 15, describe drive behavior estimating device of the present invention, drive supporting device in detail, reach the first suitable embodiment in the vehicle evaluating system.

The summary of (1) first embodiment

In the present embodiment, by with everyone different drive behavior characteristic modelization of driver, carry out vehicle control or drive supporting with driver's characteristic coupling, support and feel at ease and comfortable safe driving.In addition, constitute the vehicle design evaluation system that has used based on the objective appraisal benchmark of statistical data.

At this, as long as driver's modelling is handled and the output computing that uses a model is simple, such use just can be easily and realization at an easy rate

Therefore, in the present embodiment, by using GMM (gauss hybrid models) in driver's model, can generate driver's model of each each people of driver easily, and then, have the calculating of the probability of condition by maximization, easily infer and export the driver behavior action.

Promptly, in the drive behavior estimating device of present embodiment, the running data that will be made of various features amounts such as accelerator operation amount, the speed of a motor vehicle, vehicle headways is as learning data, the gauss hybrid models that will utilize EM (expectation value maximization, Expectation Maximization) algorithm to calculate adopts as driver's model.

This gauss hybrid models parameter of probability density function when calculating by the EM algorithm that probability density distribution obtains simultaneously constitutes, and generates by each driver and then the characteristic quantity keeping that scope is used etc. and infer with, vehicle headway with, brake service by driver's accelerator operation as required.

Also have, by measure the running data Y except specific characteristic quantity x in a plurality of characteristic quantities that use in driver's model (=y1, y2 ...), calculate with respect to the maximum posterior probability in driver's model of this running data Y, infer characteristic quantity x.

For example, generate driver's model of driver's first in advance, carry out the automatic cruising (ACC) of following front vehicles and travelling automatically.

That is, in ACC, detect running data Y such as the speed of a motor vehicle except characteristic quantity x=accelerator operation amount or vehicle headway, calculate the maximum posterior probability in driver's model of first.This value is estimated to be the accelerator operation amount of driver's first possibility practical operation under identical conditions, carries out accelerator control (engine throttle control) according to the accelerator operation amount of inferring.

Thus, the driver behavior near generating the driver that driver's model is arranged carries out accelerator operation.

In addition, when travelling, use the driver's model that generates in advance, infer the characteristic quantity of various drive behaviors, estimate the performance of vehicle thus based on the imaginary space of track model at the vehicle that makes certain design value data (performance data).

(2) first embodiments in detail

Fig. 1 be expression with based on the generation of driver's model of the drive behavior estimating device in the present embodiment with based on the figure that infers relevant principle of the drive behavior of driver's model of generation.

Also have, about the generation of driver's model and inferring of drive behavior, explanation as characteristic quantity use vehicle velocity V, with the vehicle headway F of front vehicles and these behavioral characteristics amount Δ V, a Δ F (differential value), secondary behavioral characteristics amount Δ Δ V, Δ Δ F (second differential value), with driver's model as accelerator operation, use accelerator operation amount G and a behavioral characteristics amount Δ G, as driver's model of brake service, use the situation of a brake service amount B and a behavioral characteristics amount Δ B.

In the drive behavior estimating device of present embodiment, will speed up running data 1 that device operational ton, the speed of a motor vehicle, vehicle headway etc. constitute as learning data, and utilize the EM algorithm generate in advance based on driver's model 2 of cooresponding each driver's of running data GMM.

Also have, at the drive behavior of inferring the driver (for example, accelerator operation amount) under the situation, use cooresponding driver's model 2, calculating to the measured value of the running data 1 of moment t (V, F, Δ V ...) 3 maximum posterior probability 4, infer the accelerator operation amount 5 of this driver's possible operation thus.

In this routine drive behavior estimating device, the driver is based on following supposition, that is: based on the current speed of a motor vehicle, vehicle headway, and these once, secondary behavioral characteristics amount, determine the operational ton of accelerator pedal and brake pedal.

Below, describe the generation of driver's model and the principle of inferring of drive behavior in detail.

(A) study of driver's model

In having used driver's model 2 of GMM, need learning data, use running data 1 as characteristic quantity.

Running data 1 uses the time series data every the measuring interval s of regulation (s be any, but s=0.1 second) in the present embodiment.

Running data 1 is the data that in fact driver of the object that generates as driver's model drives, and by using the running data 1 of measuring in advance, preserving, can carry out the study of off line.In addition, in fact also can use the running data 1 of The real time measure, collection when the driver drives.

In the drive behavior estimating device of present embodiment,, can realize the modelling of mating with each driver's characteristic by each driver is generated GMM.

As the characteristic quantity (running data 1) of driver's model, as mentioned above, use the speed of a motor vehicle, vehicle headway, and these once, secondary behavioral characteristics amount and accelerator operation amount, an and behavioral characteristics amount of accelerator pedal operation amount.

Like this, the modelling by add the behavioral characteristics amount to characteristic quantity, the time relationship before and after considering can access level and smooth and naturality is high infers the result.

Also have, in explanation, illustrated to use once to reach the situation of secondary behavioral characteristics amount, but only used a behavioral characteristics amount also can.

Equally, also can realize driver's modelling about brake pedal.

Also have, use with, vehicle headway scope with, brake pedal etc. under the situation of a plurality of driver's models generating accelerator pedal, the data beyond accelerator pedal operation amount, the brake pedal operational ton etc. (V, F, Δ V, Δ F ...) use identical data also can.

In the present embodiment, about the behavioral characteristics amount in the running data, obtain by calculating by the measured value of accelerator operation amount, the speed of a motor vehicle, vehicle headway, but practical measurement also can.

Also have, in the present embodiment, distribute (GMM), generate driver's model 2 by the mixed Gaussian that calculates running data 1.

That is, probability density distribution when using the EM algorithm to calculate, parameter={ the λ i of probability density function when calculating for running data 1,-μ i, ∑ i|i=1,2,3 ... M} is as based on driver's model 2 of GMM, is stored in the storing mechanism such as data bank.

At this, λ i represents weighting ,-μ i represents the mean vector group, and ∑ i represents to disperse to disperse the ranks group jointly, and M represents mixed number.In addition, as-μ i, in front the expression have-the symbolic representation vector.

Like this, in the GMM of present embodiment, also consider the association between the feature dimension, use full-shape to disperse ranks jointly.

Also have, as the EM algorithm, for example, (electronic information communication association 1988, P51～P54) carry out inferring of the EM algorithm that distributes based on mixed Gaussian according to the holy work in middle river " based on the voice understanding of probability model ".

(b) drive behavior (accelerator pedal and brake pedal operational ton) infers

The driver with based on the current speed of a motor vehicle, vehicle headway, and these once, secondary behavioral characteristics amount determine accelerator pedal and brake pedal this hypothesis of operational ton as the basis, infer the drive behaviors such as operational ton of pedal.

That is, distribute in the time of by characteristic quantity, infer drive behavior in the highest accelerator pedal operation amount of the condition lower probability of giving etc.

Be the maximized problem that has the probability of condition like this, utilize the calculating of maximum posterior probability.

That is, accelerator pedal operation amount Λ G (t) and brake pedal operational ton Λ B (t) are under the condition of giving y (t), infer the value x (t) of the probability maximum that makes the condition of having, and as maximum posterior probability, use following formula (1), (2) to calculate respectively.

∧ G (t)=arg max p (G| Δ G, V (t), F (t), Δ V (t), Δ F (t), Δ Δ V (t), Δ Δ F (t)) formula (1)

∧ B (t)=arg max p (B| Δ B, V (t), F (t), Δ V (t), Δ F (t), Δ Δ V (t), Δ Δ F (t)) formula (2)

At this, as Λ G (t), the front represents that it is presumed value that the symbol of Λ is arranged.

p(G|ΔG，V，F，ΔV，ΔF，ΔΔV，ΔΔF)

＝{p(G，V，F，ΔV，ΔF，ΔΔV，ΔΔF，ΔG)}/{∫∫…∫p(G，V，F，ΔV，ΔF，ΔΔV，ΔΔF，ΔG)dΔG，dV，dF，dΔV，dΔF，dΔΔV，dΔΔF}

p(B|ΔB，V，F，ΔV，ΔF，ΔΔV，ΔΔF)

＝{p(B，V，F，ΔV，ΔF，ΔΔV，ΔΔF，ΔB)}/{∫∫…∫p(B，V，F，ΔV，ΔF，ΔΔV，ΔΔF，ΔB)dΔB，dV，dF，dΔV，dΔF，dΔΔV，dΔΔF}

In formula (1), (2), t represents that constantly G, B, V, F, Δ are represented accelerator pedal operation amount, brake pedal operational ton, the speed of a motor vehicle, vehicle headway respectively, reached the behavioral characteristics amount.

Wherein, the value that makes the accelerator pedal of probability maximum of the condition of having and brake pedal by in minimum value in peaked interval, at certain scale spacing (for example, be 100 scales till 0 to 10000) carry out numerical integration, calculate probability thus, the accelerator pedal during with this probability maximum and the value of brake pedal also can as inferring the result.

Fig. 2 represents and the relevant summary of inferring based on the drive behavior of maximum posterior probability.

In this Fig. 2, for simply, be illustrated in the situation that certain constantly infers Λ x (t) during the characteristic quantity y (t) of t of having given.

Fig. 3 represents the structure of drive behavior estimating device.

Drive behavior estimating device in the present embodiment possesses: driver's model generating unit 10; Drive behavior estimating portion 11.

This drive behavior estimating device is realized by the computer system that possesses CPU, ROM, RAM etc.

Also have,,, also can form the structure that does not possess driver's model generating unit 10 by using the driver's model that generates by other devices as drive behavior estimating device.

Driver's model generating unit 10 possesses: driver's information acquisition portion 101; Running data acquisition portion 102; While probability density distribution calculating part 103; Probability density function parameter storage part 104 simultaneously.

The driver model of driver's information acquisition portion 101 by being used to set up generation and the information of driver's corresponding relation constitute driver ID.That is, be the driver ID of the driver when determine measuring the running data that obtains by running data acquisition portion 102.

Running data acquisition portion 102 obtains as the running data that is used to generate based on the learning data of driver's model of GMM.

Fig. 4 is illustrated in running data acquisition portion 102 and obtains running data.

As shown in Figure 4, as running data, have running environment data (a), reach driver's service data (b).

But it can may not all be data necessary as the data of data use that these running datas have been enumerated.According to the driver's model that generates, suitably select data.

Shown in Fig. 4 (a), the running environment data have travel condition data and condition of road surface data.

Travel condition data is according to travelling or the data of environmental change, has the speed of a motor vehicle, vehicle headway, weather, blocking up has or not (degree of blocking up), brightness, other data.

The condition of road surface data are data of state of expression road, are not according to the data of environmental change.The condition of road surface data have road category, the form of mating formation, road width, lane number, friction coefficient, concavo-convex coefficient, bending curvature, inclined-plane, inclination, direct-view, other data.

Shown in Fig. 4 (b), driver's service data has bearing circle operational ton, accelerator pedal operation amount, brake pedal operational ton, vehicle headway and keeps scope amount, other data.It is in the majority that this driver's service data becomes the situation of the drive behavior (presumed value of characteristic quantity x) that driver's model of use generating infers.Therefore, driver's service data of the cooresponding number of number of the driver's model that obtains and generate.For example, use under the situation of accelerator mockup with driver's model and driver behavior, obtain accelerator pedal operation amount and brake pedal operational ton in the operation of generation accelerator pedal.

In addition, under the situation that generates the shared driver's model of accelerator operation and brake service, also obtain both.

Concerning the running data that obtains in running data acquisition portion 102, can obtain the running data measuring in advance and preserve in the lump, in addition, the data that the a/s sample time obtains to detect successively during the actual driving of driver also can.

Simultaneously probability density distribution calculating part 103 (Fig. 3) with the running data that obtains as learning data, probability density distribution when calculating in the gauss hybrid models.

Probability density function parameter { λ i when will obtain based on the result calculated of this while probability density distribution calculating part 103,-μ i, ∑ i} and the driver ID that obtains in driver's information acquisition portion 101 set up corresponding relation, and it is stored in probability density function parameter storage part 104 simultaneously.

Also have, in the time of storage the probability density function parameter save as whose can be distinguished (driver ID), to driver's model of what (drive behavior of inferring).

Fig. 5 is expression generates processing based on driver's model of driver's model generating unit 10 of such formation a diagram of circuit.

(step 10) obtains running data (step 11) in the lump or successively in running data acquisition portion 102 to obtain driver's information in driver's information acquisition portion 101.

Also have, the order of step 10 and step 11 can be opposite, also can be arranged side by side.

Secondly, at the same time in the probability density distribution calculating part 103, to obtain running data as learning data, calculate probability density distribution (step 12) simultaneously, the probability density function parameter is as driver's model simultaneously, itself and driver's information are set up corresponding relation, it is stored in probability density function parameter storage part 104 (step 13), end process simultaneously.

In Fig. 3, drive behavior estimating portion 11 possesses: driver's information acquisition portion 111; Running data acquisition portion 112; Driver's Model Selection portion 113; Maximum posterior probability calculating part 114; And the presumed value efferent 115 of characteristic quantity X.

111 acquisitions of driver's information acquisition portion are used for the driver ID of the object of definite driver's model.

This driver's information spinner will obtain by driver's (driver or other operators) input.

Also have, driver's (driver) body weight or height and the information that can determine the driver as driver's information, and are set up corresponding relation with it with driver ID and preserved, obtain driver's information, definite thus driver ID also can.

Running data acquisition portion 112 obtains: in the running data (N kind characteristic quantity) that when driver's model that will use is generated by driver's model generating unit 10, uses except the running data (N-a kind of characteristic quantity) of the drive behavior (characteristic quantity x) of inferring by this driver's model.

Driver's Model Selection portion 113 selects suitable driver's model (probability density function parameter simultaneously) based at the driver ID of driver's information acquisition portion 111 acquisitions and the running data that obtains in running data acquisition portion 112 from while probability density function parameter storage part 104.

Maximum posterior probability calculating part 114 will be applicable at the running data that running data acquisition portion 112 obtains and use above-mentioned formula (1), (2) etc. to calculate maximum posterior probability by driver's model of selecting in driver's Model Selection portion 113.

The presumed value efferent 115 of characteristic quantity X will be exported as the presumed value of characteristic quantity x in the value that maximum posterior probability calculating part 114 calculates.

Fig. 6 is that expression uses the driver's model that generates to infer the diagram of circuit of the processing of specific drive behavior.

At first, obtain driver's information (step 20) in driver's information acquisition portion 111.

Also have, in running data acquisition portion 112, the running data (step 21) when obtaining current time point (t constantly).At the running data of this acquisition is N-a kind of running data except characteristic quantity x.

Also have, the order of step 20 and step 21 can be opposite, also can concurrent processing.

Also have,, select cooresponding while probability density function parameter (driver's model) and read in (step 22) from while probability density function parameter storage part 104 according to driver's information and running data.

Also have,, do not use running data, from driver's information, select driver's model also can in order to select driver's model.In this case, before the acquisition of running data, select driver's model also can.

In addition, the device by using drive behavior estimating device, for example sometimes, the pilot instrument (ACC device) of following automatically described later is selected driver's model in advance, and under the sort of situation, step 20 and step 22 can be omitted as required.

Secondly, in maximum posterior probability calculating part 114, the running data that obtains is applicable to driver's model of selection calculates maximum posterior probability (step 23).

About the calculating of maximum posterior probability, if the drive behavior (characteristic quantity x) of inferring then utilizes above-mentioned formula (1), if the brake service amount is then utilized formula (2) for accelerator operation amount.

Also have, the result of calculation in the maximum posterior probability calculating part 114 that calculates is exported (step 24) from the presumed value efferent 115 of characteristic quantity X as the presumed value based on the characteristic quantity x the moment t of its driver's model, return main program.

(3) drive supporting device

Secondly, the application examples to the drive behavior estimating device that used above explanation is that drive supporting device describes.

This drive supporting device is to follow front vehicles, carries out the device of automatic follow running (ACC).In this drive supporting device, by using driver's model by the running data generation of the driver in driving, vehicle headway when ACC is moved, accelerator operation, brake service are operated automatically, carry out ACC, eliminate driver's human discomfort near the sensation of travelling of driver oneself driving.

(4) summary of drive supporting device

Custom when driving, the driver is arranged respectively.In service at ACC in the past, be that steady state value travels automatically in order to keep the speed of a motor vehicle or vehicle headway merely, therefore, when regulating the speed of a motor vehicle, vehicle headway or with front vehicles apart from the time the using method of accelerator drg different with the custom of own (driver), the problem of human discomfort is arrived in cenesthesia.

In the drive supporting device of present embodiment, the driver behavior of the reality by the driver under the state that can move ACC generates driver's model with running data in advance as learning data.Thus, the custom of these drivers' such as operational ton of accelerator the when relation of the reflection speed of a motor vehicle and vehicle headway or distance adjustment or drg driver behavior in the driver's model that generates.

That is,, will keep the relevant custom of operation as driver's model learning and preservation with driver's workshop by driver's common driving.

The driver's model that generates will from car data (accelerator operation amount, brake service amount, the speed of a motor vehicle ...), the front vehicles data (vehicle headway, relative velocity, car type ...), road environment data (situation that mixes of brightness on every side, direct-view, raceway groove, moisture conditions, road surface μ, lane width, road, other) these three information opening relationships and generating.

Except the geography information of periphery, also comprise the information that changes among the TPO such as the situation that mixes of brightness on every side or weather condition of road surface, road in the road environment data.

Brightness, rainfall situation around these are inferred by table (constantly) or head lamp switch, wiper switch etc.

In addition, carry lux meter or rainfall sensor and road surface μ detecting sensor, lane means of identification, the various sensors of observing the situation that mixes on every side, pattern recognition device etc., initiatively obtain around the car type etc. of front vehicles information also can, obtain weather forecast or VICS information etc. from the networking.

Also have, if carry out ACC (automatically in the follow running), then by monitoring similarly with common ACC, keep and the vehicle headway of front vehicles that the driver is liberated from accelerator operation.

By with this presumed value as the basis, keep the scope amount of inferring along vehicle headway and carry out vehicle headway and regulate, along the accelerator operation amount of inferring control engine throttle, drg, carry out the automatic follow running of vehicle.

At this, vehicle headway is kept the scope amount of inferring and is meant: by the driver in being similar to the scene of present situation, usually and front vehicles between the vehicle headway kept calculate, reflect the scope of the vehicle headway in this scene that the driver has a liking for thus.

The accelerator operation amount of inferring is meant: the accelerator operation of carrying out when usually regulating vehicle headway in the driver is similar to the scene of present situation is calculated, thus, the way of the distance the during distance of shortening under the described scene of reflection driver hobby and front vehicles (catch up with rapidly, slowly catch up with etc.).

The brake service amount of inferring is meant: the brake service of carrying out when usually regulating vehicle headway in the driver is similar to the scene of present situation is calculated, thus, the way of the distance during reserving distance with front vehicles and shorten under the described scene of reflection driver hobby (reserve rapidly, slowly reserve etc.).

Like this, the operation of the custom by reproducing the driver is carried out vehicle headway and is kept, and can reduce the human discomfort that the driver feels.

In addition, also can reproduce for around brightness or the custom of the employing method of the vehicle headway that changes of external essential factor such as weather, condition of road surface, form more system near driver's perception.

(5) drive behavior estimating device is detailed

Fig. 7 has represented to be suitable for the structure of the drive supporting device of drive behavior estimating device.

Drive supporting device possesses: running data acquisition portion 12, accelerator portion 13, drg portion 14, ECU (electronic control package) 15, ACC switch 16, homing advice 17.

Also have,, do not need them whole about utilizing the structure of the drive supporting device that Fig. 7 illustrates, each one or the device that can use in order to carry out automatic follow running is illustrated, can be according to the function of the drive supporting device that adopts etc., suitably select, constitute drive supporting device.

Running data acquisition portion 12 possesses: detect the car speed sensor 120 from the speed of a motor vehicle of vehicle; The vehicle headway sensor 121 of the vehicle headway of detection and front vehicles; The camera head 122 in the place ahead of shooting vehicle; GPS+ car inter-vehicle communication range finding portion 123; Road environment information collection component 124.

GPS+ car inter-vehicle communication range finding portion 123 usefulness GPS devices determine from the position of vehicle (latitude, longitude), and, by with the car inter-vehicle communication of front vehicles, accept the coordinate figure (latitude, longitude) of front vehicles, calculate two vehicle headways between vehicle thus.

Road environment information collection component 124 is collected road environment information from head lamp switch, wiper switch, road surface μ detecting sensor, lane means of identification, VICS, vehicle-surroundings monitoring sensor, other devices, each one.

Accelerator portion 13 possesses: accelerator pedal 131; Accelerator pedal position sensor 132; Engine's throttling control valve device 133.

Drg portion 14 possesses: brake pedal 141; Brake pedal position sensor 142; Arrester control device 143; Stoplight stop lamp 144.

ECU15 possesses: front vehicles identification tracking part 151; ACC could judging part 152; Driver's model generating unit 153; ACC handling part 154.

ACC handling part 154 possesses: vehicle headway monitors keeps the 154a of portion; The 154b of accelerator operation portion; The 154c of brake service portion; The 154d of drive behavior estimating portion.

In the 154d of drive behavior estimating portion, calculate each presumed value that vehicle headway is kept scope, accelerator operation, brake service.

ECU15 comprise possess CPU, the computer system of each one of ROM, RAM, interface constitutes.

ACC switch 16 is that the driver selects whether to carry out the switch that ACC travels.

For under the situation of closing, generate driver's model at this ACC switch 16.

In addition, under the situation about opening, use the driver's model that generates, infer the driver behavior amount, carry out automatic follow running corresponding to the amount of inferring at ACC switch 16.

Homing advice 17 possesses: current location test section 171; Cartographic information 172.

Also have, current location test section 171 utilizes the current location (latitude, longitude) of detection vehicles such as gps signal receiving device, and brings into play function as the GPS of running data acquisition portion 12.

Below, the action in the drive supporting device that constitutes like this is described.

Fig. 8 represents that ACC handles with the automatic generation of driver's model.

At first, ECU15 judges that whether ACC switch 16 is for cutting out (step 31).

At ACC switch 16 for opening (that is step 31; Under the situation not), require ACC to travel, therefore, do not generate driver's model, monitor the situation that ACC switch 16 cuts out by the driver.

On the other hand, be (step 31 under the situation of closing at ACC switch 16; Be), can ECU15 judges discern front vehicles (step 32).

About the identification of front vehicles, identification in the vehicle identification tracking part 151 forwardly, and detect under the situation with the distance of front vehicles with vehicle headway sensor 121, be judged as and can discern.The identification of the front vehicles in the front vehicles identification tracking part 151, follow the trail of by being undertaken by the photographed images of the front vehicles of camera head 122 shootings.

(step 32 under the situation that can not discern front vehicles; Not), can not get the vehicle headway data, can not generate ACC driver's model, therefore, return step 31.

On the other hand, (step 32 under the situation that can discern front vehicles; Be), then, ECU15 judges whether to carrying out the scene of ACC operation by the cartographic information of homing advice 17.For example, under the situation in express highway or capital high speed uplink are sailed, do not have the interval (Inter of predetermined distance to close い on toll road or the outside route) under the situation of the travels down of the line that is open to traffic, be judged as in the travels down that can carry out the ACC operation.

ECU15 be not the scene (step 33 that can carry out the ACC operation; Under the situation not), return step 31, reprocessing.

On the other hand, so long as can carry out the scene (step 33 of ACC operation; Be), ECU15 carries out driver's model illustrated in fig. 5 and generates processing (step 34), end process.

Generate in the processing at this driver's model, generate vehicle headway and keep scope driver's model, accelerator operation driver's model, brake service driver's model.

Like this, in the drive supporting device of present embodiment, as long as open ACC switch 16, then carry out the driving condition of the reality in the environment that ACC travels based on reality, automatically generate driver's model, therefore, the custom of the described drivers' such as operational ton of accelerator the when relation of driver's model of the generation reflection speed of a motor vehicle and vehicle headway or adjustable range or drg driver behavior.

Below, use each the driver's model that generates as described above, the operation of the situation that the ACC that reality is carried out moves describes.

Fig. 9 is the diagram of circuit of the action of expression ACC processing.

ECU15 monitors ACC switch 16 whether opened (step 41).

If detect (the step 41 of opening of ACC switch 16; Be), ECU15 collects road environment information (step 42), at this, the road environment information that ECU15 collects is to generate the running data of handling in the running data that uses in (step 34) except the operational ton of inferring (accelerator operation amount, brake service amount, reach vehicle headway keep the scope amount of inferring) at driver's model.

Secondly, in the vehicle identification tracking part 151, can judgement discern the front vehicles (step 43) that should follow to ECU15 forwardly.

Under the situation that can not discern front vehicles (step 43: not), can not keep automatic follow running, therefore, open ACC switch 16, and, should look like and inform driver's (step 44), end process by sound or image.

On the other hand, if can discern front vehicles (step 43; Be), then ECU15 carries out drive behavior estimating illustrated in fig. 6 and handles (step 45), calculates vehicle headway and keeps the scope amount of inferring, the accelerator operation amount of inferring, the brake service amount of inferring.

Also have, judge current vehicle headway (step 46), as if in the vehicle headway setting range, the ECU15 accelerator opening (step 37) of maintaining the statusquo then.

On the other hand, if vehicle headway is below the vehicle headway setting range, then ECU15 is according to the presumed value of the drg of inferring in driver's model in brake service, control brake device control setup 143 (steps 48).

In addition, if vehicle headway is more than the vehicle headway setting range, then ECU15 controls engine's throttling control valve device 133 (steps 49) according to the presumed value of the accelerator of inferring in driver's model in accelerator operation.

Then, ECU15 judges whether ACC switch 16 has closed (step 50), if do not close (step 50; Not), then return step 42, continue ACC and drive, if close (step 50; Be), end process then.

Like this according to the drive supporting device of present embodiment, driver's model of the custom of the accelerator when generation has reflected the relation of the speed of a motor vehicle and vehicle headway or adjustable range or its drivers' such as operational ton of drg driver behavior, based on the accelerator operation amount of inferring, brake service amount, vehicle headway setting range by this driver's model, operation accelerator (engine throttle) or drg, therefore, realization is near the ACC of driver's the sensation of travelling.

(6) vehicle evaluating system

Secondly, be that vehicle evaluating system describes to second application examples of having used the drive behavior estimating device that illustrates.

This vehicle evaluating system is under various road holdings or driving conditions, and the running data during based on a plurality of driver's actual travel such as professional driver or average driver generates driver's model.

Also have, in this vehicle evaluating system, be not the performance evaluation of carrying out vehicle by actual steering vehicle, but the imagination that will use driver's simulator to launch is travelled and is operated acceleration capability, deceleration performance, the maneuvering performance of estimating vehicle, stablizes various projects such as rideability by the presumed value (operational ton of bearing circle operational ton, accelerator pedal, the operational ton of brake pedal etc.) of using the driver's model that generates.

(7) vehicle evaluating system is detailed

Figure 10 represents the summary of vehicle evaluating system.

This vehicle evaluating system shown in Figure 10 possesses: the driver's model 19 based on GMM that is used to infer driver's amount of pedal operation; Calculate vehicle power calculating part 20 based on the amount of pedal operation of inferring from the acceleration/accel a of car (t); Use is from the acceleration/accel a of car (t) and go ahead of the rest the position renewal speed of a motor vehicle v (t) of vehicle and the running environment renewal portion 21 of vehicle headway F (t); Calculate speed of a motor vehicle v (t) and vehicle headway F (t) once, the behavioral characteristics amount calculating part 22 of secondary variable quantity (behavioral characteristics amount).

The running data that driver's model 19 uses by using actual travel or driver's simulator etc. to measure, be considered as (for example by professional driver or average driver etc. as the driver of the targeted customer's of the vehicle of evaluation object total collection, 100 examples) running data generates a plurality of driver's models.

As driver's model, generate accelerator with driver's model 192 and drg with driver's model 193, judge and which selected to be suitable for wherein with being suitable for judging part 191.

Also have, except accelerator uses driver's model 192 and drg with driver's model 193, generation is used to infer bearing circle driver's model of driver's bearing circle manipulated variable, is made as with suitable judging part 191 and can also selects the driver to use driver's model.

Driver's model 19 is that drive behavior estimating device is suitable for, and accepts the speed of a motor vehicle or so-called characteristic quantity of vehicle headway and behavioral characteristics amount, the part of the accelerator pedal of appointment driver possible operation and the value of brake pedal.In this driver's model 19, as the explanation of carrying out with drive behavior estimating device, the driver will based on the current speed of a motor vehicle, vehicle headway, and these once, secondary behavioral characteristics amount, the hypothesis of operational ton of determining accelerator pedal and brake pedal is as the basis.

In vehicle power calculating part 20, by certain vehicle velocity V (t-1) in the accelerator pedal operation amount G (t) during t and brake pedal operational ton B (t), previous moment of reaching constantly, use auto model (with as vehicle performance datas such as the relevant vehicle weight of the vehicle of evaluation object, engine performance, brake performance, gear compare), calculate and add the speed of a motor vehicle.

In auto model, MATLAB (based on FORTRAN, machine language that can powerful processing array computing) has been installed with reference to not the model pinched of the driving simulator of including that is used in learning data.

In this Vehicular system, consider the friction coefficient of gear ratio or car weight, road etc., calculate the speed of a motor vehicle that adds of vehicle.

In running environment renewal portion 21, the acceleration/accel a (t) when using the moment t from 20 inputs of vehicle power calculating part calculates vehicle velocity V (t+1), the vehicle headway F (t+1) of the ensuing moment (t+1), and upgrades.

Speed of a motor vehicle during the ensuing moment (t+1) and vehicle headway are by following calculating.

V(t+1)＝V(t)+a(t)×T

F(t+1)＝Df(t+1)—(Dm(t)+V(t+1)×T)

Wherein, a (t) represents the speed of a motor vehicle that adds from 20 outputs of vehicle power calculating part, the operating range of the front vehicles till Df (t) the expression moment t, the operating range from vehicle till Dm (t) the expression moment t.

In addition, T is the update time (sample cycle) of system, is T=0.1 second in the present embodiment.

In addition, in order to calculate vehicle headway, by each operating range constantly time of driving earlier, operating range when obtaining the moment t of described driving earlier and poor from the operating range of car are calculated vehicle headway thus.

Figure 11 represents the structure of drive behavior estimating device.

As shown in figure 11, drive behavior estimating device comprises: be used in the data of evaluation, the execution portion and the evaluation result data of evaluation.

As service contamination, use vehicle performance data 25, evaluation driver's model 26, simulation to carry out with data 27, track model 28, as the evaluation result data, export and store the evaluating data 36 that travels.

Execution portion as estimating possesses: road data expansion portion 29; Running environment expansion portion 30; The behavior estimating portion 31 of travelling; Vehicle power calculating part 32; Simulate the handling part 34 that travels; Rideability evaluation portion 35, these utilize the computer system that is made of CPU, ROM, RAM etc. to constitute.

Vehicle performance data 25 are the performance datas as the vehicle of evaluation object, shown in Figure 12 (a), are made of each data such as vehicle weight, engine performance, brake performance, gear ratio, suspension spring constants.

Estimate with driver's model 26 and use each the driver's model that in above-mentioned drive behavior estimating device, generates.

Simulation is carried out with data 27 shown in Figure 12 (b), is the travel condition data of launching on the imaginary space, is made of the speed of a motor vehicle, vehicle headway, weather, having or not etc. of blocking up.This travel condition data is used and is pressed t=t1, t2, t3 constantly ... data as the data of sequential.

Track model 28 shown in Figure 12 (c) is and the relevant data of launching on imaginary space of test travel.

If comparative evaluation execution portion and drive behavior estimating device illustrated in fig. 10, the behavior estimating portion 31 of then travelling is equivalent to driver's model 19, vehicle power calculating part 32 is equivalent to vehicle power calculating part 20, and travel handling part 34 and running environment expansion portion 30 of simulation is equivalent to running environment renewal portion 21 and behavioral characteristics amount calculating part 22.

Secondly, the design evaluation that the vehicle in the vehicle evaluating system that constitutes like this is described is handled.

Figure 13 represents to design the diagram of circuit of estimating the action of handling.

Figure launches (step 62) from vehicle performance data 25 input vehicle power calculating parts 32 (steps 61) with the vehicle power model in virtual space.

Then, to road data expansion portion 29 input track model 28 (steps 63), 30 input simulations are carried out with data 27 (step 64) to running environment expansion portion, launch simulation execution environment (step 65) at virtual space thus.

Also have, estimate with driver's model 26, begin to import the execution (step 67) that simulation is travelled from t=0 to behavior estimating portion 31 inputs of travelling.

Then, calculate driver's behavior estimating value (accelerator pedal operation amount G (t) and brake pedal operational ton B (t)) (step 68) in the running environment data (running data) of the behavior estimating portion 31 of travelling during by moment t.

Also have, in vehicle power calculating part 32, accelerator pedal operation amount G (t) during by moment t and brake pedal operational ton B (t), and the vehicle velocity V (t-1) in the previous moment, use gear than or the vehicle performance datas 25 such as friction coefficient of car weight, road, calculate acceleration/accel a vehicles such as (t) estimation data 33 of travelling.

The vehicle ' estimation data of calculating 33 except acceleration/accel a (t), still shown in Figure 12 (d) from vehicle speed, vehicle headway, center-of-gravity position, tire angle, yaw-rate, pitching rate etc.

Also have, in handling part 34 was travelled in simulation, the vehicle ' estimation data 33 when using the moment t that is calculated by vehicle power calculating part 32 was calculated vehicle velocity V (t+1), the vehicle headway F (t+1) of the ensuing moment (t+1), and upgraded (step 70).

In addition, with simulating the vehicle ' estimation data 33 (step 71) that the handling part 34 that travels calculates t=t+1.

Also have, the simulation execution environment (step 72) in running environment expansion portion 30 by vehicle ' estimation data 33 renewal t=t+1, storage (step 73) driving trace to road data calculates in performance evaluation portion 35 travelling.

Also have, judge to carry out with the simulation of the total data till the moment tn of data 27 processing of travelling and whether finishes (step 74), if there is not end (step 74 about simulation; Not), then return step 68, continued to use the simulation of driver's model.

(step 74 under the situation that the processing till the moment tn has finished; Be), export the evaluating data that travels, end process (step 75) by rideability evaluation portion 35.

As the evaluating data that travels from 35 outputs of rideability evaluation portion, shown in Figure 12 (e), as the acceleration chart of acceleration capability output to accelerator opening, as the retardation curve of deceleration performance output to the brake service amount, as the travel curve of maneuvering performance output to the bearing circle operational ton, as stable rideability, output is to the driving trace of road direction etc.

(8) simulated experiment

(8-1) is based on the study of driver's model of GMM

For the study of GMM, use driving simulator, include the driving data.

Road is a straight line, has variation in order to make learning data, adopts the trace data of driving earlier, so that all speed of a motor vehicle occur.

In addition, carry out twice 10 minutes travel, with twice amount as learning data.

The trace of driving a vehicle earlier shown in Figure 14 (a), the driving data of including shown in Figure 14 (b).

Owing to considered the variation of the trace of first driving, therefore as can be known, all speed of a motor vehicle occur.The model learning separately of accelerator pedal operation and brake service is to have the 16 regular distributions of Multidimensional and Hybrid (GMM) that mix that full-shape disperses ranks jointly.

(8-2) analog result and investigation

In order to estimate the vehicle evaluating system of constructing, the trace of preparing not comprise in the learning data of driving is earlier included it.

Road is a straight line, and the trace of driving uses the trace of including in actual environment earlier.The trace of driving earlier that is used to estimate shown in Figure 14 (c) and uses.

Use the data of this elder generation's driving, generate drive behavior, relatively actual driving data.Simulated conditions is as described below.

Learning data; 20 minutes (twice 10 minutes)

Characteristic quantity; V, F, G, Δ V, Δ G, Δ Δ V, Δ Δ F

Road; Straight line

The Δ window width; 0.8 second

Mixed number; 16

Update time; 0.1 second

Figure 15 is illustrated in the analog result of having used the vehicle evaluating system of present embodiment under the above condition.

Figure 15 is result (b), and the result (c) of accelerator pedal of result (a), the vehicle headway of the speed of a motor vehicle, and solid line is an analog result, and dotted line is represented actual data.

As shown in figure 15, for example, think and catch the feature of the waveform of actual driver behavior signal well, carried out modelling well based on GMM about the accelerator pedal operation.

More than, first embodiment in drive behavior estimating device of the present invention, drive supporting device, the vehicle evaluating system is illustrated, but the present invention is not limited to the embodiment of explanation, carries out various changes in the scope that can put down in writing in each request scope.

For example, in the embodiment of explanation, learn the relation of the speed of a motor vehicle or so-called signal of vehicle headway and drive behavior signal by learning data based on driver's model of GMM, therefore, if become the condition (end of distribution) that does not have in the learning data, then can not carry out inferring of amount of pedal operation well.

For example, in dig tracking, surpass travelling in the vehicle headway of 100m or the so-called vehicle headway of 1m and be not included in the learning data, under the situation that becomes the condition that is not contained in such learning data, can not infer its result well, continue away from first driving, or collide.

Therefore, the situation for fear of so also can be set as follows, that is: vehicle headway be below the L1 (for example, 2m is following) time, do not carry out applying full application of brake based on the inferring of driver's model, at vehicle headway is (for example, more than 100 under) the situation, to will speed up device and trample to standard-sized sheet more than the L2.

Secondly, with reference to Figure 16～Figure 36, describe second embodiment that is fit in driver's model generating apparatus of the present invention and the drive supporting device in detail.

The summary of (9) second embodiments

In the present embodiment, by detecting driver's Biont information, whether identification is driver's usual state.Also have, in driver's driving, collect driving condition data (from information of vehicles, for example, accelerator, drg, the operational ton that turns to, the speed of a motor vehicle, vehicle headway, acceleration/accel etc.), extract the part that the driver in these driving condition data drives with usual state, generate driver's model.

Thus, can under the situation that the driver does not recognize, generate driver's model just often automatically.

In addition, only situation about will drive with normal condition based on driver's Biont information is as just often drive behavior, and therefore generation driver model, can form driver's model of more high-precision neutral gear.

In the present embodiment, for example, picture is on the highway in one way three tracks, at signal is that green point of crossing is under the situation of right-hand corner special vehicle diatom right-hand corner, the subtend car is arranged, cross walk has under pedestrian's the situation the same, generates driver's model by each scene (situation) from the vehicle-surroundings environment in travelling.

In addition, drive behavior just often of being inferred by the driver's model that generates by real-time comparison and current driver's drive behavior monitors whether current driver's drive behavior is " by usually ", or omission (Left takes off).

" common " driven and the index during current driving as a comparison, for example, and use driver's " speed of response " and " shakiness ".

And then, in the present embodiment, not only based on the variation of driver's model evaluation drive behavior, but also add the variation of Biont information, the information of driver's state is represented in compound thus judgement, detects the reduction of driver's fatigue or attention more accurately.

Its result having under the situation of omission from driver's drive behavior originally, notes prompting, warning or information indicating to this, can carry out thus supporting with the safe driving of this people's coupling.

In addition, the driver condition in the reduction omen stage before of finding tangible fatigue or attention can be detected, the guiding of rest equal altitudes can be before fatigue peaks, urged in advance.

In the present embodiment, with first embodiment in the same manner, use GMM (gauss hybrid models) in driver's model, can generate driver's model easily by each people of each driver thus, and then, have the calculating of the probability of condition by maximization, easily infer and export the driver behavior action.

Promptly, driver's model generating apparatus of present embodiment, drive supporting device, and the drive behavior decision maker in, will be with the learning data that constitutes by various features amounts such as accelerator operation amount, the operational ton that turns to, vehicle headway, acceleration/accels as learning data, the gauss hybrid models that utilizes EM (ExpectationMaximization) algorithm to calculate adopts as driver's model.

The parameter of probability density function constituted when this gauss hybrid models obtained by utilizing the EM algorithm to calculate the while probability density distribution, as required, the characteristic quantity of keeping that scope is used etc. and inferring with, vehicle headway with, brake service by each people of each driver and driver's accelerator operation generates.

Also have, measure running data Y in a plurality of characteristic quantities that are used in driver's model except specific characteristic quantity x (=y1, y2 ...), calculate with respect to the maximum posterior probability in driver's model of this running data Y, infer characteristic quantity x thus.

For example, use the driver's model in the situation identical with the running environment (situation) of vehicle-surroundings, to driver's model input current from the car state, (for example infer its later driving condition, characteristic quantity x=accelerator operation amount) time changes, compare with the driving condition of reality, shakiness of the delay of decision operation or operation etc. has or not thus.

(10) second embodiments in detail

Figure 16 represents to use the structure of the drive supporting device of driver's model generating apparatus.

Drive supporting device possesses: ECU (electronic control package) 50; From information of vehicles acquisition portion 51; From vehicle-surroundings environmental information acquisition portion 52; Biont information acquisition portion 53; Information provides portion 54; Driver's model handling part 55; Data store 56.

Also have, about utilizing the structure of the drive supporting device that Figure 16 illustrates, it all not necessarily, to the generation that can be used in the driver's model that carries out in the present embodiment and drive supporting and each one or the device that use be illustrated, can constitute drive supporting device according to the suitable selections such as function of the drive supporting device that adopts, in addition, can append other units that use has same function.

ECU50 is made of the computer system of each one that possesses CPU, ROM, RAM, interface.

ECU50 carries out: based on the indication that portion 54 is provided from the supervision of driver's drive behavior of the acquired information of information of vehicles acquisition portion 51, based on the supervision of driver's Biont information of the acquired information of Biont information acquisition portion 53, to the information as driver's auxiliary content of drive supporting.Again that generation, the output of driver's model is the required data of ECU50 are supplied with to driver's model handling part 55.

Possess from information of vehicles acquisition portion 51: bearing circle is handled angle transducer 511, accelerator pedal position sensor 512, brake pedal position sensor 513, speed gauge 514, acceleration pick-up 515, electronic situation acquisition portion 516, timer 517, other sensors done.

Among Figure 17 illustration obtain from information of vehicles acquisition portion 51 as driver's operation information from information of vehicles.

As shown in figure 17, bearing circle is handled angle transducer 511 and is detected bearing circle operational ton (angle), and accelerator pedal position sensor 512 detects accelerator operation amount, and brake pedal position sensor 513 detects the brake service amount, and speed gauge 514 detects the speed of a motor vehicle.

Acceleration pick-up 515 detects yaw axis acceleration/accel, axis of pitch acceleration/accel, roll shaft acceleration/accel.

Flash light operation conditions, lamp operation conditions, Windshield Wiper operation conditions detect in the electronic situation acquisition portion 516 that does.

Timer 517 is measured the various times such as driving the moment, driving time.

Possess from vehicle-surroundings environmental information acquisition portion 52: vehicle-surroundings information acquisition portion 521, road information acquisition portion 522, and Network Dept. 523.

Vehicle-surroundings information acquisition portion 521 possesses: various sensors such as infradred sensor, millimeter wave sensor, ultrasonic transduter, pattern recognition device, vehicle headway sensor.The image processing of image outside the car that pattern recognition device carries out making a video recording in image-input device, there are object in obstacle or pedestrian, the vehicle etc. of identification vehicle-surroundings.

Among Figure 18 illustration the vehicle-surroundings environmental information that in vehicle-surroundings information acquisition portion 521, obtains.

As shown in Figure 18, utilize vehicle-surroundings information acquisition portion 521 to obtain vehicle, pedestrian, obstacle, other various information.

As the information of concrete acquisition, for example, the kind (car, motor bike, bicycle etc.) of the vehicle that the periphery that obtains to detect by each vehicle exists, vehicle headway, relative velocity, attribute (subtend car, also driving, (left and right) car etc. of keeping straight on.

Equally, for pedestrian, obstacle, also obtain information respectively to it.

Road information acquisition portion 522 possesses: the GPS device or be used to that detects the current location of vehicle obtains and the cartographic information that peripheral information such as has or not of the current location road corresponding information that detects or signal.

In addition, road information acquisition portion 522 possesses the pattern recognition device of identification marking or road environment, but the image recognition of this pattern recognition device and vehicle-surroundings information acquisition portion 521 is total.

Among Figure 19 illustration the vehicle-surroundings environmental information that obtains in road information acquisition portion 522.

In road information acquisition portion 522, as shown in figure 19, obtain road category, road shape, road width, having or not and state, road attribute (traffic regulation), other various information from the brightness in car position, condition of road surface, road, signal.

Traffic information net such as Network Dept. 523 and VICS or weather information sensor are connected, and obtain traffic information or weather information.

Among Figure 20 illustration the vehicle-surroundings environmental information that in Network Dept. 523, obtains.

As shown in figure 20, at acquisition congestion informations such as VICS there be having or not of the having or not of the having or not of the distance of blocking up, the distance that mixes, accident, block off traffic, chain restriction etc.

In addition, the weather information that obtains in the weather information sensor has Weather information, precipitation probability, temperature, other information such as fine, cloudy, rain.

Obtain from vehicle-surroundings environmental information acquisition portion 52 from the vehicle-surroundings environmental information with above-mentioned from information of vehicles acquisition portion 51 obtain from the part of information of vehicles (for example, based on information such as directly the advancing of bearing circle operational ton, right-hand corner, turnon lefts) together, according to status list 563, be used in the setting of situation.

Biont information acquisition portion 53 obtains to be used for to judge that the driver of the driving of vehicle still is the Biont information of error state for normal condition, as the sensor that is used for it, possess ecg scanning device, blood-pressure gage, heart rate sensor, sweating sensor, reach other sensors.

Heart rate and diaphoretic volume detect with specific time interval in Biont information acquisition portion 53 under the situation that vehicle begins to travel, supply with to ECU50.

Heart rate sensor for example is disposed at the electrode of bearing circle by utilization, the hand getting of the driver from drive detects heart rate with heart rate signal.Also have, heart rate sensor can be with the sensor configuration of special use on drivers' such as wrist health.

The sweating sensor configuration is on bearing circle, by the change-detection sweating state according to sweating state value of current flowing.

Among Figure 21 illustration the Biont information that obtains in Biont information acquisition portion 53.

In Biont information acquisition portion 53, with heart potential, R-R interval, heart rate, Respiration Rate, body temperature, blood pressure, skin potential, dehydration component, myoelectric potential, E.E.G current potential etc. as object.

Information provides portion 54 to possess: the driver behavior assisted parts, voice output portion, the picture efferent that are used to carry out or warning auxiliary with the cooresponding driver behavior of driver's driving condition.

Among Figure 22 illustration provide the information that portion 54 provides, auxiliary content by information.

As shown in Figure 22, in the driver behavior assisted parts, as revising auxiliary based on driver's driver behavior, the operation of travel direction dish is auxiliary, driver behavior is auxiliary, brake service is auxiliary, therefore, controls the output of the torque value of each operating portion.For example, in that operation has under the unstable situation based on driver's bearing circle, bearing circle becomes the important place and carries out the torque operation, under the situation a little less than the pedal force of drg, the output of the tread-on quantity of drg is become the earth assist.

In addition, according to driver's state, voice output portion output warning voice, the picture efferent shows the warning picture.

Driver's model handling part 55 possesses: driver's model generating unit 551, driver's model storage part 552, driver's model efferent 553.

Driver's model generating unit 551 is as driver's model generating apparatus performance function, be stored in information of vehicles acquisition portion 51 under the situation that driver's the state in information of vehicles that obtains is a normal condition from information of vehicles, generate driver's model from this normal condition from information of vehicles.

Normal condition from information of vehicles by when obtaining this information, storing by the situation of determining from information of vehicles that obtains from vehicle-surroundings environmental information acquisition portion 52, generate driver's model by each situation.

Driver's model storage part 552 is kept at the driver's model that generates in driver's model generating unit 551 by each situation.

In driver's model generating unit 551, if store specified amount to each situation from information of vehicles, then generate driver's model of this situation, it is stored in driver's model storage part 552.Also have, when obtaining from information of vehicles at every turn, with driver's model of cooresponding situation with stored before from the information of vehicles merging and regenerate driver's model, renewal driver model.Also have, the renewal of driver's model is in each obtain cooresponding situation new during from information of vehicles, but generates, upgrades and also can when storing specified amount appending at every turn.

Figure 23 concept nature is represented the memory contents of driver's model storage part 552.

As shown in Figure 23, driver's model is by the situation classification.Each driver's model a, b, c of storage ... with cooresponding status data (situation a, b, c ...) line, system is as the mark performance function that is used to quote driver's model.

By such setting, when the retrieval of driver's model, can obtain system hardware (cash) operation of driver's model of the situation of " fatigue conditions of driver under certain level " in the lump.

Driver's model efferent 553 based on the cooresponding driver's model of specific situation n n, infer and export the operational ton of the driver under the normal condition, promptly, the driver behavior amount of (just often) during with respect to situation n common.

By relatively this infers driver behavior amount and current from information of vehicles, obtain being used to judge that the basic data of the state (response delay, shakiness etc.) of drive behavior described later is drive behavior missing data at interval in required time.

Also have,, driver's model storage part 552 is stored in data store 56 also can with driver's model generating unit 551 in the ECU50 realization driver model handling part 55 and two functions of driver's model efferent 553.

The driver's model that stores present embodiment in data store 56 generates and handles, reaches required various data or the table of driver behavior auxiliary process.

Bending curvature comprises: the recording medium of optically reading information such as semiconductor recording mediums such as magnetic recording medias such as soft dish, hard disk, tape, memory chip or IC-card, CD-ROM or MO, PD (phase change can wiping type CD), with the recording medium of other the whole bag of tricks record data or computer program.

Recording medium uses different media also can according to recorded content.

Store drive behavior missing data 561 in the bending curvature, from information of vehicles 562, in addition, storage condition table 563.

Drive behavior missing data 561 is the driver behavior amounts just often of being inferred by the driver's model n with respect to the situation n in the current driving and based on the differential data from the operational ton of information of vehicles of reality, calculates and preserves at interval in required time for the situation n in the current driving.

From information of vehicles 562 by each situation store when travelling with normal condition from information of vehicles.Storing this time point of specified amount, generating and the cooresponding driver's model of this situation from information of vehicles.Driver's model obtains being updated of cooresponding situation at every turn when information of vehicles after once generating.

Status list 563 be used for by obtain from information of vehicles and from the vehicle-surroundings environmental information determine cooresponding situation a, b, c ... table.

Figure 24 concept nature is represented the content of status list.

As shown in Figure 24, by with driver's model a, b, c ... cooresponding each situation a, b, c ..., set the condition flag that is used to become this situation.

With regard to condition flag, selecting data in each from information of vehicles and each small project in the vehicle-surroundings environmental information.

Secondly, illustrate based on the various processing actions of the drive supporting device of formation as described above.

Figure 25 is the diagram of circuit that driver's model of driver's model of generation driver " common drive behavior " (just often) generates the processing action of handling.

In the present embodiment, being created in the vehicle ' of driver's model carried out, but carry out driver's Biont information under steam, from information of vehicles, from the collection and the storage of vehicle-surroundings environmental information, about the setting of condition flag and the generation of driver's model, in vehicle ', carry out also can in addition.

Driver's model generating unit 551 is the Biont information (step 110) when each time point is collected by Biont information acquisition portion 53 in vehicle '.Also have, via various information such as ECU50 collection of biological body information (down together).

Secondly, driver's model generating unit 551 judges by monitoring its variable condition from the Biont information of collecting, store whether current driver's state is normal condition (step 111).

Figure 26～Figure 28 concept nature represents to judge whether the driver is the method for normal condition.

Figure 26 represents the change by driver's heart rate, the state that supervision is rocked or the impatient spirit (mental) that causes changes.

Shown in Figure 26 (a), between threshold value h1 about the regulation and h2, accompany under the situation of measurement of Heart Rate value, be judged as normal condition (stable status).

On the other hand, shown in Figure 26 (b), detecting the measurement of Heart Rate value is below the downside threshold value h1, or under the situation more than the upside threshold value h2, is judged as and rocks or the impatient unusual state (unstable state) that causes.

Also have, in the present embodiment, shown in Figure 26 (b), under the situation about departing from from the both sides between last lower threshold value h1, h2, be judged as error state at the appointed time, unusually also can but the situation that either party's threshold value exceeds schedule time is judged as.

Figure 27 represents to resolve the state that monitors that spirit changes by Electrocardiographic long-range navigation thatch topology.

In long-range navigation thatch topology is resolved, with arbitrarily constantly R-the R of the heart potential during n be made as RRn at interval, the R-R of the heart potential during with ensuing moment n+1 is made as under the situation of RRn+1 at interval, generates the value that transverse axis is got RRn, the chart of the value of the bent RRn+1 of the longitudinal axis.At this, R-R is the time gap of the peak value of heart potential to ensuing peak value at interval, is equivalent to the interval of heart rate.

Resolve as can be known according to this long-range navigation thatch topology, under the situation of strenuous status extremely, shown in Figure 27 (a), heart beat interval is a same intervals, and the set of topology point concentrates on a place of y=x line.

In addition, under suitable strenuous status (state that the attention of appropriateness is arranged), observe the swing that heart beat interval has appropriateness, shown in Figure 27 (b), being integrated on the y=x line of topology point slenderly marked and drawed.

In addition, under scatterbrained state, the swing of heart beat interval is big, and shown in Figure 27 (c), the set that observes the topology point also reaches the set of direction bulging rectangular with it to the initial point direction on the y=x line.

In addition, having under the state of sleepiness, shown in Figure 27 (d), the set of topology point is that the topology area of the y=x line direction of heart beat interval broadens, but the width of initial point side is narrow, and from along with the tendency that broadens away from initial point.

Resolve by this long-range navigation thatch topology, judge normal condition (strenuous status of appropriateness) and abnormal state of affairs (strenuous status, the scatterbrained state of extreme, the state of sleepiness is arranged).

It still is that the state of parasympathetic system advantage judges whether situation into normal condition that Figure 28 represents by the state of the Biont information that obtains by the sympathetic nervous system advantage.

As shown in Figure 28, for example, measure the size of pupil, under its size is loose big situation, be judged as sympathetic nervous system advantage the most, because strenuous status extremely might cause attention to reduce by driver's photographed images.On the contrary, the size of pupil is under the situation of contraction state, and being judged as parasympathetic nerve is the state that relaxes of advantage, and according to the degree of shrinking, attention may be low, and attention may be extremely low.

On the other hand, be under the situation of pupil size of appropriate advantage at sympathetic nerve, be judged as the high normal condition of attention under the strenuous status of appropriateness.

By each mensuration project of the effect project shown in Figure 28 (heart rate, cardiac contractile force etc.) of the size that comprises pupil, pre-determine the value of the anxiety that is used to divide into anxiety extremely, appropriateness, the state that relaxes (attention is low), this one of four states of the state that relaxes (attention substrate).

Also have, in the generation of driver's model, whether judge normal condition, but monitor that at driver's Biont information described later driver's Biont information situation of handling in (with reference to Figure 33) judges in (step 142), based on the method that illustrates among Figure 26～Figure 28 judge strenuous status extremely, appropriateness strenuous status, scatterbrained state, the state that relaxes, these five states of state of sleepiness are arranged.

As mentioned above, judge by Biont information whether the driver is normal condition, if normal (step 111; Be), then driver's model generating unit 551 from from information of vehicles acquisition portion 51 and from vehicle-surroundings environmental information acquisition portion 52 collect just often from information of vehicles and generate the information (step 112) of usefulness as normal driving person model from the vehicle-surroundings environmental information.

Among Figure 29 illustration when the point of crossing right-hand corner by from information of vehicles acquisition portion 51 and from vehicle-surroundings environmental information acquisition portion 52 obtain from information of vehicles with from the vehicle-surroundings environmental information.

Under the situation of as shown in Figure 29 point of crossing right-hand corner, as the information that obtains, obtain road category, condition of road surface, from vehicle speed, from the car position, the having or not of the having or not of working direction, kind, front vehicles relative position, front vehicles relative velocity, subtend car, subtend car car type, subtend car relative position, subtend car relative velocity, pedestrian, pedestrian's kind, pedestrian's position, pedestrian's working direction, weather etc. from the having or not of the state of the signal of car side (red, green, yellow etc.), front vehicles, front vehicles.

Also have, in the present embodiment, obtain these information, be used in the setting of situation described later, but may not necessarily need use them whole, carry out also can based on the setting of the situation of the information of arbitrary part, on the contrary, carry out also can based on the setting of the situation of more detailed information.

Driver's model generating unit 551 by collect from information of vehicles with from the vehicle-surroundings environmental information, according to status list 563 (with reference to Figure 24), set condition flag, with collect just often from information of vehicles be stored in respective conditions in information of vehicles 562 (step 113).

Secondly, driver's model generating unit 551 according to collect and store from information of vehicles 562, generate and the cooresponding person's model (step 114) of just often using the normal driving of the situation of setting in step 113, return main program.

On the other hand, judge whether to be normal condition, if normal (step 111 by Biont information; Not), then driver's model generating unit 551 with just often in the same manner, generate the information of use as abnormal driving person model, when unusual from information of vehicles and from the vehicle-surroundings environmental information from from information of vehicles acquisition portion 51 with from 52 collections (step 115) of vehicle-surroundings environmental information acquisition portion.

Driver's model generating unit 551 is set condition flag from information of vehicles with from the vehicle-surroundings environmental information according to status list 563 (according to Figure 24) by what collect, with collect unusual the time from information of vehicles be stored in respective conditions from information of vehicles 562 (step 116).

Secondly, driver's model generating unit 551 when collecting store unusual from information of vehicles 562, generate with the situation of setting in step 116 cooresponding when unusual with abnormal driving person's model (step 117), return main program.

Also have; in the present embodiment; to being illustrated according to the situation of Biont information for normally still generate the normal driving unusually person's model and abnormal driving person model; but for example; as error state, the last lower threshold value that the Biont information ratio is stipulated high and lowly generates the driver's model corresponding to the state of Biont information respectively.

At this, the generation based on driver's model of driver's model generating unit 551 is described.

In the present embodiment, utilize GMM to generate driver's model.

About based on the generation of driver's model just often of the drive supporting device (driver's model generating apparatus) of present embodiment and based on the principle of inferring of the driver behavior amount of the driver's model that generates as described in the explanation in the Fig. 1 and first embodiment.

Also have, about characteristic quantity, the combination of other information in all information that use obtains in information of vehicles acquisition portion 51 generates also can.

In the drive supporting device of present embodiment, each running data 1 (from information of vehicles) of each situation that will be made of accelerator operation amount, the speed of a motor vehicle, vehicle headway etc. is as learning data, utilize the EM algorithm generate in advance based on driver's model 2 of the cooresponding GMM of each situation.

Also have, driver's model generates by each people of driver also can.

Also have, at the drive behavior of inferring the driver (for example, the driver behavior amount) under the situation, use cooresponding driver's model 2, the measured value of the running data 1 when calculating t constantly (V, F, Δ V ...) 3 maximum posterior probability 4, infer the accelerator operation amount 5 of this driver's possible operation thus.

Each operational ton of inferring is like this inferred ensuing operational ton as running data, compare, calculate drive behavior missing data thus with each each iede measured value (from information of vehicles) of the moment.

In this routine drive supporting device, the driver will based on the current speed of a motor vehicle, vehicle headway, and these once, secondary behavioral characteristics amount determines that the hypothesis of operational ton of accelerator pedal and brake pedal is as the basis.

Below, the principle of inferring of the generation of driver's model and drive behavior as in first embodiment explanation as described in.

Also have, (A) in the study of driver's model, running data 1 is the data as the actual driving of driver of the object of driver's model generation, uses the running data 1 of The real time measure, collection when the actual driving of driver.In addition, by using the running data 1 of measuring in advance and storing, carry out the study of off line.

In addition, about with the relevant summary of inferring based on the drive behavior of maximum posterior probability, as shown in Fig. 2 of explanation in the first embodiment.

To using the driver's model that generates by each situation as described above, driver's drive behavior of state of determining driver's drive behavior monitors to handle and describes.

Figure 30 is the diagram of circuit that expression driver drive behavior monitors the processing action of handling.

ECU50 is from collecting from information of vehicles with from vehicle-surroundings environmental information (step 120) from information of vehicles acquisition portion 51 with from vehicle-surroundings environmental information acquisition portion 52.

Secondly, ECU50 shown in Figure 31 (a), based on obtain from information of vehicles with from the vehicle-surroundings environmental information, set condition flag (step 121).

Also have, ECU50 carries out matching treatment based on the condition flag of setting with status list 563, and retrieval is fit to the situation from present situations such as vehicle-surroundings environmental informations of acquisition, judges that thus whether cooresponding driver's model exists (step 122).

(step 122 under the non-existent situation of cooresponding driver's model; ), do not return main program.

Shown in Figure 31 (b), detect the situation that is fit to present situation on the other hand, (step 122 under the situation that cooresponding driver's model exists; Be), ECU50 reads driver's model with the situation line that is fit to from driver's model storage part 552, with it to 553 outputs (step 123) of driver's model efferent.

Secondly, ECU50 will by obtain during from information of vehicles acquisition portion 51 at moment t from information of vehicles (measured value) as initial value (t), to 553 inputs (step 124) of driver's model efferent.Like this, importing to driver's model from information of vehicles (t) when driver's model efferent 553 is incited somebody to action moment t calculated maximum posterior probability, thus the presumed value " t+1 " (step 125) of the drive behavior data (operational ton) when ECU50 output time t+1.

Then, when ECU50 obtains current (t+1 constantly) from information of vehicles (t+1) (step 126), the drive behavior missing data (" t+1 "-(t+1)) when calculating t+1 constantly is stored in drive behavior missing data 561 (step 127) with it.

Also have, ECU50 judges whether the drive behavior missing data 561 of storage has stored stated number (step 128), if stated number is less than (step 128; Not), then the presumed value " t+1 " of the operational ton that will infer in step 125, shifts to step 125 to driver's model input (step 129) as (t), thus, and then continues the storage (step 125～127) of drive behavior missing data 561 constantly.

On the other hand, if store the drive behavior missing data 561 (steps 128 of stated number; Be), then ECU50 judges drive behavior omission tendency by the state of drive behavior missing data 561, and output (step 130), returns main program.

In the present embodiment, as the omission of drive behavior tendency, to " delay of speed of response " have or not and these two projects that have or not of " shakiness of operation " are judged.

The relatively presumed value of the driver behavior amount (common driving) under the normal condition of driver's model efferent 553 output and the operational ton (from information of vehicles) of current driving of concept nature among Figure 32.

In this Figure 32, import the initial value of current driver behavior amount to driver's model from driver behavior amount (common driving) expression of driver's model output, the driver just often as the output valve of the highest operational ton of the common drive behavior probability of getting, if expression just often then should be able to be carried out the operational ton of the hypothesis of such driving (should be operational ton) usually.

For this imaginary operational ton, contrast the operational ton of current driving, the tendency of the delay of judgement speed of response and the shakiness of operation.

For example, shown in Figure 32 (a), under the situation that the operational ton of being inferred by driver's model increases along with effluxion, at the appointed time after the warp, by obtain to be judged as the delay tendency that has speed of response from information of vehicles acquisition portion 51 based under the situation of the operational ton of information of vehicles.

In addition, shown in Figure 32 (b), compare with the operational ton of in driver's model, inferring, increase along with effluxion based on the operational ton that obtains from information of vehicles, or under the situation about reducing, increase, reduction (absolute value of drive behavior missing data) are under the above situation of specified value, are judged as driver behavior and exist unstable.

On the other hand, the operational ton of inferring by driver's model with based on obtain under the roughly consistent situation of the operational ton of information of vehicles, the absolute value that is drive behavior missing data is under the situation of the following state continuance of specified value, be judged as all do not have response delay, unstable normal condition.

Secondly, according to the diagram of circuit of Figure 33, illustrate that the supervision of the Biont information of the driver in travelling is handled.

At first, ECU50 is collected, is stored the Biont information (step 141) of each time point by Biont information acquisition portion 53 in vehicle '.

Secondly, ECU50 is by monitoring its variable condition by the Biont information of collecting, store, utilize with Figure 26～Figure 28 in the identical method of method that illustrates, judge and the state (situation) (step 142) of current driver's Biont information return main program.

Also have, more than the driver driver Biont information of explanation monitors that handling driver's model that can be also used as in the travelling of illustrating among Figure 25 generates driver's Biont information in handling and collect whether the variation of (step 110) and driver's Biont information is normal judgement (step 111).

In this case, generate in the processing at driver's model, monitor in step 111 whether driver's state is normal condition, this, as described in explanation in Figure 25, judging yet, export according to Figure 26～Figure 28 is not to be equivalent to any state under the situation of normal condition.

Also have, under the situation that dual-purpose should be judged, any of driver's model generating unit 551, ECU50 carried out its judgement.

Bending curvature shown in Figure 34 (a) and (b), by the state of driver's eyes, judges that normal condition, sleepiness, fatigue state also can.

By the judgement of the state of these eyes driver's model generate handle (step 111), and driver's information monitoring handle to use among the either party of (step 142) or the both sides and also can.

Specifically, shown in Figure 34 (a),, utilize image processing to detect the moving of aperture, sight line of number of winks, time nictation, eyelid, judge the state of sleepiness according to its value or state as driver's state.

In addition, shown in Figure 34 (b), under the situation that number of winks increases or under the situation of the mobile spasm of eyelid, under the situation about rubbing one's eyes, under the situation at the angle that rubs one's eyes, be judged as fatigue.

The diagram of circuit of Figure 35 action that to be expression handle based on the drive supporting of driver's driving condition and Biont information situation.

In ECU50, state as the driver, acquisition driver drive behavior monitors the drive behavior omission tendency (step 151) of judging in the step 130 of handling (Figure 30) and exporting, and, obtain driver's Biont information and monitor driver's Biont information situation (step 152) of judging and export in the step 142 of handling (Figure 33).

Also have, ECU50 omits tendency and Biont information situation by the drive behavior that obtains, and determines the content of drive supporting, carries out the method (step 153) to driver's support, returns main program.

The content of the driver's that will be inferred by drive behavior that obtains and Biont information situation among Figure 36 state (a) and the drive supporting that carries out corresponding to the driver's who infers state has generated table.

Also have, the table of Figure 36 (b) is stored among the ROM of ECU50.

Shown in Figure 36 (a), concentrations, fatigue, careless driving, sleepiness, impatience, so-called driver's such as gaze around state are gradually inferred in the combination of each drive behavior (reaction delays, unstable, reaction is relaxed+unstable, do not have both) and each Biont information situation (strenuous status, scatterbrained state, the state that relaxes of strenuous status extremely, appropriateness, the state of sleepiness is arranged).

Also have, corresponding to each state of inferring by these each combinations, shown in Figure 36 (b), ECU50 provides portion 54 to carry out reminding, providing the facilities information suggestion to have a rest, note prompting+α 1, provide information to eliminate drive supportings such as atmosphere based on the attention of voice or vibration by information.

Also have, in the drive supporting shown in Figure 36, ECU50 is for+α 1 and since concentrate on drive beyond and dangerous, therefore, urge and note reminding and increase with the control of the vehicle headway of front vehicles automatically etc.

In addition, ECU50 is for+α 2, note reminding in order to concentrate on driving, and, carry out following action, that is: use the careless more driving of interactive function or sensor investigation driver more what to be felt load, guiding is used to take out one rest, or carry out worry and eliminate discussion, or improve the problem that the driver entertains.

In addition, ECU50 is for+α 3, and the attention that wakes eyes up provides, and, take guiding of having a rest etc. rapidly.

In addition, in ECU50, as the content of note reminding, carry out driver behavior or Biont information what cause problem and specific description that warning etc. make the driver understand easily also can.

More than, the embodiment in driver's model generating apparatus of the present invention and the drive supporting device is illustrated, but the present invention is not limited to the embodiment of explanation, in each request item, can carry out various distortion in the scope of record.

For example, in the embodiment of explanation, determine that by drive behavior and the Biont information situation judged the situation of the content of drive supporting is illustrated, but determine that by the drive behavior of judging the drive supporting content also can.

Claims (15)

1. drive behavior estimating device is characterized in that possessing:

Driver's model, it travels the time series data of detected N kind characteristic quantity as learning data with escort vehicle, and describes the summary that each data exists in the N dimension space and distribute;

Characteristic quantity obtains mechanism, and it obtains at least a above characteristic quantity except specific characteristic quantity x in the described N kind;

Maximum posterior probability is calculated mechanism, and it is calculated with respect to the maximum posterior probability in described driver's model of the characteristic quantity of described acquisition;

Output mechanism, it is based on described maximum posterior probability of calculating, and output is with respect to the presumed value of the described specific characteristic quantity x of the characteristic quantity of described acquisition.

2. drive behavior estimating device according to claim 1 is characterized in that,

Described N kind characteristic quantity comprises: for the n kind (time variation amount of the characteristic quantity of n＜N).

3. drive behavior estimating device according to claim 1 and 2 is characterized in that,

Described characteristic quantity x comprises: the operational ton of the handling device of driver's direct control and the time variation amount of this operational ton.

4. according to each described drive behavior estimating device in the claim 1～3, it is characterized in that,

Described driver's model as learning data, is described as the probability distribution that each data exists the time series data of described N kind characteristic quantity by the GMM that utilizes the EM algorithm to calculate (gauss hybrid models).

5. drive supporting device is characterized in that possessing:

Each described drive behavior estimating device in the claim 1～4, its use: as characteristic quantity, used accelerator operation amount, brake service amount, from the speed of a motor vehicle of vehicle, with the accelerator of the vehicle headway of front vehicles with driver's model and drg driver's model, and infer accelerator operation amount and brake service amount as described characteristic quantity x;

Running data obtains mechanism, and it obtains the speed of a motor vehicle and vehicle headway from vehicle;

Ride control mechanism, it is for the running data of described acquisition, according to the accelerator operation amount of being inferred by described drive behavior estimating device, and brake service amount, control engine throttle and brake pedal carry out the automatic follow running to described front vehicles thus.

6. vehicle evaluating system is characterized in that possessing:

Each described drive behavior estimating device in the claim 1～4, its use: as characteristic quantity, used accelerator operation amount, brake service amount, from the speed of a motor vehicle of vehicle, with the accelerator of the vehicle headway of front vehicles with driver's model and drg driver's model, and infer accelerator operation amount and brake service amount as described characteristic quantity x;

Acquisition is as the mechanism of the vehicle performance data of the vehicle of evaluation object;

Obtain the mechanism of simulation with running data and track model;

The vehicle power computing mechanism, it is for being applicable to that by running data and track model with described acquisition described drive behavior estimating device obtains accelerator operation amount and brake service amount, inferring the trace that comprises as the acceleration/accel of the vehicle of described evaluation object;

Assessing mechanism, it estimates the rideability as the vehicle of described evaluation object by the trace of described estimated vehicle.

7. driver's model generating apparatus is characterized in that possessing:

The state decision mechanism, it judges driver's state;

Driver behavior information acquisition mechanism, it obtains the driver behavior information in the vehicle ';

Driver's model generates mechanism, and it is based on the driver behavior information of described acquisition, makes the driver's model with driver's the cooresponding driver behavior of state.

8. driver's model generating apparatus according to claim 7 is characterized in that,

Described state decision mechanism judges at least whether driver's state is normal.

9. according to claim 7 or 8 described driver's model generating apparatus, it is characterized in that,

Running environment obtains mechanism, and it detects specific running environment;

Each running environment is stored described driver behavior information,

Described driver's model generates mechanism, and each makes driver's model by described running environment.

10. according to each described driver's model generating apparatus in the claim 7～9, it is characterized in that,

Possess: Biont information obtains mechanism, and it obtains driver's Biont information,

Described state decision mechanism is judged driver's state based on the Biont information of described acquisition.

11. a drive supporting device is characterized in that possessing:

Driver's model obtains mechanism, and it obtains driver's model of the driver behavior under the normal condition;

The driver behavior estimating mechanism, it uses driver's model of described acquisition, infers the driver behavior of driving usually under normal condition;

Drive behavior decision mechanism, it judges driver's drive behavior by described driver behavior of inferring with based on the driver behavior of current driver behavior information;

Drive supporting mechanism, it carries out and the cooresponding drive supporting of the drive behavior of described judgement.

12. drive supporting device according to claim 11 is characterized in that,

Described driver's model obtains the driver model of mechanism by the driver behavior under the normal condition that generates by each driving environment, obtains and the current cooresponding driver's model of running environment.

13. according to claim 11 or 12 described drive supporting devices, it is characterized in that,

Possess: driver condition decision mechanism, it judges driver's state by driver's Biont information,

Described drive supporting mechanism carries out and the drive behavior of described judgement and the cooresponding drive supporting of driver condition of described judgement.

14. according to each described drive supporting device in the claim 11～13, it is characterized in that,

Described drive supporting mechanism is according to judging content, utilizes attention prompting, the information of voice or image to provide, at least a above drive supporting in the guiding in vibration, Rest area.

15. a drive behavior decision maker is characterized in that possessing:

Driver's model obtains mechanism, and it obtains driver's model of the driver behavior under the normal condition;

The driver behavior estimating mechanism, it uses driver's model of described acquisition, infers the driver behavior of driving usually under normal condition;

Drive behavior decision mechanism, it judges driver's drive behavior by described driver behavior of inferring with based on the driver behavior of current driver behavior information.

Images