CN113335293B - Highway road surface detection system of drive-by-wire chassis - Google Patents

Abstract

The invention discloses an expressway pavement detection system with a drive-by-wire chassis, which is used for automatically detecting pavement parameters through an unmanned aerial vehicle and a detection vehicle equipped with the drive-by-wire chassis under the condition of closed expressways, aiming at the problems that vehicles are inconvenient and dangerous to pass and the road unsealing time is difficult to determine under the condition of closed expressways. The method of managing vehicles is to predict road passability by integrating the conditions of vehicles and drivers, and further determine whether to release vehicles and plan a driving plan. In addition, the vehicle detects the road surface parameters of the current road in the driving process, predicts the passability of the road in real time, corrects the driving scheme, guides the driver to drive in a voice mode, improves the safe driving capacity of the driver, ensures that the vehicle can safely pass through, quickly recovers the traffic, and reduces the inconvenience in passing caused by road closure.

Description

A highway pavement detection system with a wire-controlled chassis

Technical field:

The invention relates to the field of traffic accident prevention, in particular to a highway road surface detection system and a vehicle management method with a wire-controlled chassis, which are used for actively detecting road surface conditions when the expressway is closed, and predicting road conditions based on vehicle and driver conditions. Passability, and then decide whether to release the vehicle and plan the driving plan. The vehicle detects the pavement parameters of the current road during the driving process, predicts the passability of the road in real time and corrects the driving plan, and guides the driver to drive in the form of voice. Vigilance to prevent traffic accidents.

Background technique:

When the expressway is closed due to snow, fog, rain and other conditions, the expressway is difficult to pass, easy to be blocked, a large number of vehicles are stranded and waiting, and the passage time is often vague, and the vehicles to be passed cannot be evacuated in time, which greatly reduces the traffic flow of the expressway. Affect the convenience of transportation.

At present, among the measures for highway driving safety or reducing traffic accidents, most of them use intelligent driving assistance systems to improve driving safety, and intelligent driving assistance systems require more sensor information for environmental recognition, which is complicated and time-consuming to process. Mature assisted driving technologies are mainly related to longitudinal control, such as lane keeping, adaptive cruise, etc., which are applicable to few traffic scenarios, and their functions are not perfect. They are temporarily unable to cope with complex traffic environments and have limited effect on improving driving safety. If the occurrence rate of traffic accidents can be predicted, that is, the road safety is evaluated, corresponding measures can be taken to remind the driver, so that the driver can take safety measures in time, which can help avoid the occurrence of traffic accidents to a certain extent, and can be easily applied to various traffic scenarios. , but there are few reports on the means of predicting and preventing traffic accidents at the overall level of the road. In addition, for road closures caused by poor expressway adhesion conditions, the existing solution is generally to wait until the bad weather completely dissipates, which is long and unreasonable. Regulations for driving do not take into account the particularity of the road surface and the special circumstances of some vehicles. Therefore, the present invention is proposed. The present invention detects road parameters on closed expressways by dispatching unmanned aerial vehicles and detection teams equipped with wire-controlled chassis, comprehensively predicts road passability according to road conditions and vehicle and driver status, and plans routes for vehicles that are allowed to pass. , broadcast the route in real time during the driving process to improve the safe passing rate and speed up the dredging of vehicles; and predict the passability of the road in real time according to the current road during the driving process, and issue the corresponding warning information to remind the driver and improve the driver's vigilance to ensure driving safety and further reduce the traffic accident rate.

Invention content:

The purpose of the present invention is to overcome the problems existing in the prior art, in the case of highway road closures caused by poor road adhesion conditions, dispatch a detection team equipped with aviation drones and wired control chassis to detect the surface parameters of the entire highway, Return the road parameter data to the vehicle management intelligent platform, and the vehicle management intelligent platform will generate the entire road surface condition distribution library, and judge the safety passing level of the vehicle based on the driver information and vehicle information, so as to decide whether to release and formulate a safe driving route for the vehicle to be released. The scheme can help to avoid traffic accidents, improve the efficiency of expressway passing, and dredge vehicles safely and quickly.

In order to achieve the above object, the present invention is realized according to the following technical solutions:

A highway road surface detection system with a wire-controlled chassis includes a road parameter detection module, a wireless communication module, a vehicle management intelligent platform, a vehicle information storage module, a voice prompt module, a driver information input module and a vehicle-mounted prediction module; wherein, the vehicle information The storage module, the voice prompt module, the driver information input module and the on-board prediction module are installed on the vehicle. The on-board prediction module includes the on-board road detection sub-module and the passability prediction sub-module; the wireless communication module includes the on-board wireless transceiver unit and the smart platform wireless module. Transceiver unit; the road parameter detection module is responsible for collecting road parameter information and sending it to the vehicle management intelligent platform; the driver information input module obtains the driver information, the vehicle information storage module collects the vehicle information, and the wireless communication module sends the driver information and vehicle information to the Vehicle management intelligent platform; vehicle management intelligent platform judges whether the vehicle can be released according to the above information, and formulates a safe driving plan for the vehicles that are allowed to be released, and sends the safe driving plan to the voice prompt module of the vehicle through the wireless communication module; the voice guidance module plays relevant The voice guides the driving route and speed of the vehicle throughout the whole process; during the driving process of the vehicle, the on-board detection sub-module in the on-board prediction module is responsible for real-time detection of road parameters, and the driver information input module and the vehicle information storage module respectively input the driver information and vehicle information. The passability prediction sub-module, the vehicle management intelligent platform sends the road parameters and safe area prediction information to the vehicle-mounted prediction module through the wireless communication module. lower, and then determine whether it is safe to pass, store the information on whether the pass is successful or not in the data storage sub-module, and send the information to the vehicle management platform at the exit of the expressway.

The road parameter detection module described in the technical solution includes a detection vehicle and an aerial drone; wherein, the detection vehicle is an unmanned intelligent vehicle based on a wire-controlled chassis, and the wheelbase, wheelbase and body height can be changed within a certain range. During the detection process, the area with the worst adhesion conditions on the road ahead is identified through image recognition by the aerial drone, and the wheelbase and vehicle distance are adjusted accordingly to adapt to the area with the worst road conditions, and the worst adhesion coefficient is detected; For two types of rover using passenger tires and commercial tires, each is equipped with 8 wheels, four types of tires, two tires on the front, rear, left and right, arranged side by side, and the tire pressure can be dynamically adjusted, so as to be based on the type of vehicle to be passed. Load to equip the rover's tires and adjust the tire pressure. There are two types of passenger tires, longitudinal pattern radial tires and longitudinal pattern ordinary bias tires. The vehicle tire flat rate statistics account for the top two. If there are no vehicles waiting, the default tires with flat rates of 65% and 55% are used; commercial tires include longitudinal pattern radial tires, transverse pattern radial tires, and mixed pattern radial tires;

The detection scheme is to send one passenger vehicle and one commercial vehicle to each lane, and drive at the specified speed of the lane. In the first detection, the detection vehicles are sent from the starting points of the two driving directions of a highway at the same time, that is, two detection vehicles. The detection vehicles of the two teams drive towards each other; the detection vehicles of the two teams upload the detected road adhesion coefficients to the vehicle management intelligent platform, and the vehicle management intelligent platform compares the adhesion conditions of the lanes in the two directions. , just send a probe vehicle in one direction to collect road parameters.

The method for obtaining road parameters by a probe vehicle in the road surface parameter detection module described in the technical solution is that the probe car has an information acquisition unit, a road curvature calculation unit, a road surface adhesion coefficient calculation unit, and a road gradient calculation unit; wherein, the information acquisition unit respectively obtains Suspension height sensor, inertial measurement unit, lidar, tire force sensor, vehicle acceleration sensor and wheel angular velocity sensor, GPS receiver signal and electronic navigation map, aerial drone takes road images 500 meters ahead of the vehicle, The road gradient and curvature are estimated through image recognition, and communicated with the wireless communication module of the probe car to send the estimated value XA0 of the longitudinal slope of the road, the estimated value of the lateral slope of the road YA0 and the estimated value of curvature C1 to the probe car; the above information collected by the information acquisition unit is used for The road curvature calculation unit, the road adhesion coefficient calculation unit, and the road gradient calculation unit are used; the road adhesion coefficient calculation unit calculates the road adhesion rate according to the vertical load of the tire and the tire force collected by the tire force sensor, and estimates the slip by the vehicle acceleration and the wheel angular velocity. Finally, according to the road adhesion rate-slip rate calibration curve, the estimated value ac of the road adhesion coefficient is obtained, and an estimated value ac of the adhesion coefficient is calculated every 5 meters; Extract the line shape of the current road position from the navigation map, and then calculate a road curvature value C2; find the road curvature design value of the corresponding road position in the road design data according to the position of the probe vehicle, and obtain the road curvature design value C3 corresponding to the current position; The road curvature design value C3, the road curvature estimated value C1 estimated by the aviation drone, and the road curvature estimated value C2 are weighted and fused to obtain the final road curvature value C. The fusion weight is set according to the weather conditions: in rain, snow, fog, hail, etc. Under bad weather conditions, the weight of the road curvature value C1 estimated by the aviation drone is 0.2, and the weight of the road curvature design value C3 is 0.5; under other meteorological conditions, the weight of the road curvature value C1 estimated by the aviation drone is 0.3 , the weight of the road curvature value C2 is 0.3, and the weight of the road curvature value C3 is 0.4; when the positioning information cannot be obtained normally, the weight of the road curvature value C1 estimated by the aviation drone is 1, and the weight of the road curvature value C2 is 0. The weight of the road curvature value C3 is 0; the information acquisition unit inputs the acceleration, wheel torque and rotational speed signals, and the point cloud data generated by the lidar into the gradient calculation subunit, and the gradient calculation subunit adopts the least square method to separate from the original acceleration sensor signal The road longitudinal gradient information is obtained, and then the road longitudinal gradient angle XA1 is obtained; the roll angle of the vehicle body relative to the chassis is estimated through the suspension height sensor information, and the road lateral gradient angle YA1 is finally estimated; the point cloud data is used to establish the Cartesian coordinate system. In the interval grid map, the normal vector of the road surface is obtained by plane fitting in the interval, and the longitudinal slope angle XA2 and the lateral slope angle YA2 of the road surface are calculated by using the normal vector; The state observer is based on the vehicle rotation The longitudinal gradient angle XA3 is estimated by the moment, rotational speed and acceleration; the vehicle lateral state observer based on the two-degree-of-freedom vehicle kinematics model estimates the road lateral gradient angle YA3 according to the front wheel angle, the yaw rate of the vehicle body, and the lateral acceleration of the vehicle body; The position of the probe car is obtained by the GPS receiver, and the design value of the slope angle is searched in the road design data according to the position of the probe car to obtain the longitudinal gradient angle XA4 and the lateral gradient angle YA4; the above five types of longitudinal gradient angles XA0, XA1, XA2 are obtained , XA3, XA4 are weighted to obtain the final longitudinal slope angle XA; the fusion weights of various longitudinal slope angles are fixed, the longitudinal slope angle XA1 fusion weight is 0.05, and the longitudinal slope angle XA1 estimated according to the acceleration sensor signal The fusion weight is 0.1, and the longitudinal observer The estimated longitudinal slope angle XA3 fusion weight is 0.2, the slope angle XA2 estimated according to lidar is 0.3, and the longitudinal slope angle XA4 fusion weight is 0.5; the above four types of lateral slope angles YA1, YA2, YA3, and YA4 are weighted and fused. , get the final side slope angle YA; when entering and leaving the side slope, the fusion weight of the side slope angle YA1 calculated based on the suspension height sensor information is 0.35, and the side slope angle YA3 estimated by the side state observer The fusion weight is 0.15. When on a slope, the fusion weight of the lateral slope angle YA3 estimated by the lateral state observer is 0.35. The fusion weight of the lateral slope angle YA1 calculated based on the suspension height sensor information is 0.15, and the lateral slope angle YA0 The fusion weight is always 0.05 for , the fusion weight for the lateral slope angle YA2 estimated from the lidar is always 0.2, and the fusion weight for the lateral slope angle YA4 is always 0.25.

The driver information input module described in the technical solution includes a human-computer interaction interface and a data storage device. Through the human-computer interaction interface, the driver inputs the basic information of the driver, the basic information of the accompanying passengers and the traffic accidents that the driver has occurred in the form of a questionnaire. Among them, the basic information of the driver includes age, gender, ID card number, driving experience, average daily driving time, physical health, mental state and occupation; the driving experience is input with natural numbers 0, 1, 2, etc. The unit is Year; the average daily driving time allows to enter an integer between 0-24, the unit is hour; there are three options for physical health status: healthy, sub-healthy, disease state, and give corresponding notes under the options for the driver to understand, "Disease state "refers to one of the symptoms of a cold, headache, and fever. In the "sub-health state", other parts of the body feel pain or discomfort but are difficult to describe. "Healthy state" refers to a normal body, no disease or pain; Options: good, poor, very poor, also with a note below the options for the driver to understand, "poor" means feeling very sleepy and difficult to concentrate, "poor" means feeling sleepy, a little Inattentiveness, "good" means no sleepiness, able to concentrate, the above description is highly subjective, input by the driver after self-assessment; career options provide two categories: professional drivers and non-professional drivers; accident characteristics include Number of accidents, accident form, accident severity, accident cause, type of vehicle involved, and accident time. Among them, accident forms include rear-end collision, scratching, and impact on fixed objects, and accident severity is divided into no casualty accident, injury accident, and fatal accident. The causes of the accident include driving in the same lane and failing to maintain a safe distance from the vehicle in front as required, improper operation, not using lights as required or driving at the specified speed in low visibility, illegally changing lanes, improper braking, violation of traffic signals, fatigue driving, illegal Driving on the road, the types of vehicles involved in the accident include heavy trucks, medium trucks, light trucks, mini trucks, large passenger cars, medium passenger cars, light passenger vehicles, small passenger cars and passenger cars. The time of the accident is specific to the date and time; information.

The vehicle information storage module described in the technical solution stores vehicle model parameters, tire information, braking system information, drive system information, steering system information, the vehicle's service life, mileage, and accident characteristics of historical traffic accidents of the vehicle. The vehicle type parameters include: Vehicle type, vehicle length, width and height, minimum turning radius, vehicle curb weight, maximum driving force, total vehicle mass, wheelbase, center of mass height, distance from center of mass to rear axle, windward area, wheelbase, wheelbase, and minimum ground clearance Clearance, approach angle, departure angle, maximum speed, maximum output torque, number of wheels and driving wheels, automatic driving level; tire information refers to tire type and tire radius; brake system information includes brake type and size parameters; drive system Information includes power type and related size parameters; steering system information includes steering system type and related size parameters; vehicle types are divided into commercial vehicles, passenger vehicles, and commercial vehicles are subdivided into heavy trucks, medium trucks, light trucks, minivans, large Passenger car, medium passenger car, light passenger car, small passenger car; the service life of the car is obtained from the driving recorder, and the time from the first time the driving recorder starts recording to the current time is the service life of the vehicle; the historical traffic of the vehicle The accident characteristics of the accident include the number of accidents, the accident form, the damage situation and the accident time. The accident form is divided into rear-end collision, scratching, and hitting the fixed object. The damage situation includes the scratched parts, damaged parts, and repaired or replaced parts. The accident time is recorded in year, month, and day.

The vehicle management intelligent platform described in the technical solution includes an environmental information acquisition module, a historical traffic accident information storage module, and a vehicle management core module; the environmental information acquisition module acquires visibility, road parameters, time and meteorological information, and the meteorological information includes rain, snow and fog. , temperature, humidity; the historical traffic accident information storage module stores traffic accident-related characteristic information, including the distribution characteristics of the vehicles involved, the time distribution characteristics of the traffic accidents, the visibility distribution characteristics of the traffic accidents, the meteorological characteristics of the traffic accidents, the location distribution characteristics of the traffic accidents, the traffic accidents Density characteristics of vehicles involved in accidents, road characteristics of traffic accidents, and characteristics of drivers involved in accidents; among them, road characteristics of traffic accidents refer to road adhesion coefficient, road curvature and slope, and characteristics of drivers involved include age, gender, driving experience, and occupation of drivers involved in accidents;

The core module of vehicle management includes a traffic flow distribution sub-unit, a safe passing grade evaluation sub-unit and a safe driving scheme planning sub-unit; among them, the traffic flow distribution sub-unit determines the number and type of vehicles to be released, and uses a neural network to estimate the current environment and road conditions. The number and type of vehicles that are allowed to be released. The neural network includes an input layer, a hidden layer and an output layer. The input layer has 4 units: time, visibility, weather vector, and road parameters. The hidden layer is divided into two layers. The first hidden layer It consists of 4 units, the second hidden layer consists of 2 units, and the output layer has 2 units. In advance, the distribution characteristics of the vehicles involved in the accident, the time distribution characteristics of traffic accidents, and the location distribution characteristics of traffic accidents in the historical traffic accident information storage module are used in advance. , traffic accident visibility distribution characteristics, traffic accident meteorological characteristics and traffic accident vehicle density characteristics to train this neural network, input time, visibility, weather, road parameter information into the neural network, and the neural network outputs the maximum number of vehicles allowed to pass and the number of vehicles allowed to pass. 's model;

The safety passing grade evaluation sub-unit evaluates the passability of the vehicle on a specific road and the driving reliability of the driver on a specific road respectively, and evaluates the safety passing grade by combining the results of these two aspects. Establish vehicle dynamics model and road model for calculation. The dynamics model includes tire model, drive system model, braking system model, vehicle body model, and air resistance model. According to the vehicle model parameters, tire information, braking model stored in the vehicle information storage module System information, driving system information, and steering system information to establish the dynamic model of the vehicle. The road model contains 4 parameters: road adhesion coefficient, curve radius of curvature, longitudinal slope and lateral slope. The data comes from the road parameter detection module; road types include The straights, curves, ramps and their combinations of different road adhesion coefficients are used to calculate the passability of the vehicle under various roads, the passable speed and transmission gear position, and obtain the position on the road where the vehicle can travel;

The driving reliability of the driver on a specific road uses the fuzzy neural network to predict the probability of the driver driving on the drivable road. The structure of the fuzzy neural network is a pre-neural network plus a fuzzy neural network, and the pre-neural network is divided into 3 sub-systems The neural network is neural network 1, neural network 2, and neural network 3. The structure of each sub-neural network is 3 layers: input layer, hidden layer, and output layer. The fuzzy neural network structure is divided into 5 layers: input layer. , fuzzy layer, fuzzy rule layer, fuzzy decision layer and output layer, the output layer of the three sub-neural networks is part of the input layer of the fuzzy neural network. The basic information of passengers and the accident characteristics of the driver's traffic accidents are input into the neural network 1 to evaluate the driver's accident probability. The accident characteristics of the driver's previous traffic accidents affect the weights of each layer of the neural network, and the neural network 1 finally Output the value between 0-100%; train the neural network 2 with the distribution data of the vehicles involved in the accident, input the model parameters of the vehicle into the neural network 2, evaluate the safety level of the vehicle, and output the probability of a traffic accident, the probability value is 0-100 %; train the neural network 3 with the traffic accident pavement feature data, and input the road parameters of the vehicle's drivable road surface into the neural network 3 in turn, and the neural network 3 predicts the probability of traffic accidents caused by these roads, and the probability value is between 0-100% ;The input layer of the fuzzy neural network contains 4 units: the driver's probability of causing an accident, the probability of the vehicle's traffic accident, the probability of a traffic accident caused by the vehicle's drivable road surface, the maximum speed of the lane limit, and the maximum speed of the lane limit is input multiple times. Vehicle speed, input the maximum speed limited by one lane at a time, the membership function of the fuzzy layer is generated by the neural network according to the historical traffic accident data, the fuzzy rules of the fuzzy rule layer are extracted from the knowledge base by the neural network, and the neural network is based on the traffic accident information. Update the real-time adjustment knowledge base, and automatically optimize the parameters of the fuzzy rules online. There are 4 fuzzy language values: {LPL, MPL, HPL, SPL}, meaning: {The safety pass level is low, the safety pass level is medium, the safety pass level is relatively high High, the safety pass level is extremely high}, the output layer outputs the safety pass level with the highest degree of membership of the road;

The safe driving scheme planning sub-unit uses the safety passing grade output from the safety passing grade to evaluate the safe passing grade of the vehicle's drivable road surface. According to the level of the safety passing grade, priority is given to connecting the road surfaces with the high safety passing grade to form a driving path with the highest comprehensive safety passing grade. ; Roads with lower safety passing grades account for no more than 5% of all road surfaces in the entire driving path, and roads with medium safety passing grades account for less than 10% of all road surfaces in the entire driving path, which are regarded as effective safe paths, and the vehicle is allowed to pass. , enter the waiting area, otherwise it will be regarded as an invalid safety path, the vehicle will not be allowed to pass, and enter the waiting area; the vehicles that can form an effective safety path will be scored according to the comprehensive safety passing grade. The scoring rule is that the safety passing grade is lower and the safety Roads with medium passing grades, high safety passing grades, and extremely high safety passing grades are respectively awarded 25 points, 50 points, 75 points, and 100 points. The total points of each road surface in the accumulated driving path are sorted according to the total, and the traffic flow allows The value releases the corresponding number of vehicles in the top ranking, and transmits the safe route data to these vehicles through wireless communication.

The in-vehicle prediction module described in the technical solution includes an in-vehicle prediction module including an in-vehicle road detection sub-module and a passability prediction sub-module. During driving, the in-vehicle road detection sub-module will detect the current road parameters, and the passability prediction sub-module. According to the detected pavement parameters and conditions, and with reference to the safety level evaluated by the platform, the current pavement is re-evaluated for passability; the vehicle-mounted pavement detection sub-module includes an information acquisition sub-unit, a pavement adhesion coefficient calculation sub-unit, and a road gradient calculation sub-unit; among them, The information acquisition unit separately acquires the signals of the suspension height sensor, the inertial measurement unit, the laser radar, the tire force sensor, the vehicle acceleration sensor and the wheel angular velocity sensor, which are used by the road adhesion coefficient calculation subunit and the road gradient calculation subunit;

The road adhesion coefficient calculation unit calculates the road adhesion rate according to the tire vertical load and the tire force collected by the tire force sensor, estimates the slip rate from the vehicle acceleration and wheel angular velocity, and finally obtains the road adhesion coefficient according to the road adhesion rate-slip rate calibration curve. Estimated value ac', detect and calculate an estimated value of adhesion coefficient ac' at the starting position of the road surface for each specific safety level;

The information acquisition subunit inputs the acceleration, wheel torque and rotational speed signals, and the point cloud data generated by the lidar into the gradient calculation subunit. The gradient calculation subunit uses the least squares method to separate the road longitudinal gradient information from the original acceleration sensor signal, and then obtains the road Longitudinal slope angle XA1'; the roll angle of the vehicle body relative to the chassis is estimated through the information of the suspension height sensor, and finally the road lateral slope angle YA1' is estimated; the point cloud data is used to establish the interval grid map in the Cartesian coordinate system, The road surface normal vector is obtained by plane fitting within the interval, and the normal vector is used to calculate the road longitudinal gradient angle XA2' and the lateral gradient angle YA2'; at the same time, the vehicle longitudinal state observer based on the vehicle longitudinal dynamics model in the gradient calculation subunit is based on the The vehicle torque, rotational speed and acceleration estimate the longitudinal gradient angle XA3'; the vehicle lateral state observer based on the two-degree-of-freedom vehicle kinematics model estimates the road lateral gradient according to the front wheel rotation angle, the vehicle body yaw rate, and the vehicle body lateral acceleration Angle YA3'; the above three types of longitudinal gradient angles XA1', XA2', XA3' are weighted and fused to obtain the final longitudinal gradient angle XA'; the fusion weights of various longitudinal gradient angles are fixed based on the acceleration sensor signals. The estimated gradient angle XA1' fusion weight is 0.2, the fusion weight of the longitudinal slope angle XA3' estimated by the longitudinal observer is 0.3, and the fusion weight of the slope angle XA2' estimated by the lidar is always 0.5; the above three types of lateral slope angles YA1', YA2', YA3' are weighted Fusion to get the final lateral slope angle YA'; when entering and leaving the lateral slope, the lateral slope angle YA1' calculated based on the information of the suspension height sensor is fused with a weight of 0.35, and the lateral slope angle estimated by the lateral state observer is 0.35. The fusion weight of the slope angle YA3' is 0.15. When on the slope, the fusion weight of the lateral slope angle YA3' estimated by the lateral state observer is 0.35, and the fusion weight of the lateral slope angle calculated based on the suspension height sensor information is 0.15. According to The YA2 fusion weight of the lateral slope angle estimated by the lidar is 0.5;

If the ratio of the difference between the road gradient and road adhesion coefficient calculated by the vehicle-mounted road surface detection sub-module and the corresponding parameter value measured by the probe vehicle and the corresponding parameter value measured by the probe vehicle exceeds 10%, the passability prediction sub-module will be activated. Re-evaluate the passability of the current road; the passability prediction sub-module uses the fuzzy neural network to predict the road passability of the model. The fuzzy neural network prediction model structure is a pre-neural network plus a fuzzy neural network, and the pre-neural network is divided into 3 The sub-neural networks are neural network 1, neural network 2, and neural network 3. The structure of each sub-neural network is 3 layers: input layer, hidden layer, and output layer. The fuzzy neural network structure is divided into 5 layers: input layer, fuzzy layer, fuzzy rule layer, fuzzy decision layer and output layer, the output layer of the three sub-neural networks is a part of the input layer of the fuzzy neural network; the neural network 1 is trained with the characteristic data of the driver involved in the accident, and the basic information of the driver and the The accident characteristics of the traffic accidents that the driver has occurred are input into the neural network 1 to evaluate the driver's accident probability, and the final output value is between 0-100%; the neural network 2 is trained with the distribution data of the accident vehicles, and the parameters of the vehicle type are input into the neural network. Network 2, evaluates the safety level of the vehicle, and outputs the probability of a traffic accident on the vehicle, the probability value is between 0-100%; trains the neural network 3 with the traffic accident road surface feature data, and inputs the collected road parameters into the neural network 3 , predict the probability of a road accident causing a traffic accident, the probability value is between 0-100%; the input layer of the fuzzy neural network contains 5 units: the driver's probability of causing an accident, the probability of the vehicle causing a traffic accident, and the road causing traffic accidents. The probability of the traffic accident caused by the environment, the speed of the vehicle, the membership function of the fuzzy layer is generated by the neural network based on the historical traffic accident data, the fuzzy rules of the fuzzy rule layer are extracted from the knowledge base by the neural network, and the neural network is based on the traffic accident information. The updated knowledge base is adjusted in real time, and the parameters of fuzzy rules are automatically optimized online. There are 4 fuzzy language values: {LPL, MPL, HPL, SPL}, meaning: {low safety pass level, medium safety pass level, safety pass level High, the safety pass level is extremely high}, the output layer outputs the safety pass level with the highest degree of membership of the road, and sends the safety pass level to the voice prompt module.

The voice prompting module described in the technical solution includes a voice playing unit and a decision-making unit; during the driving process, if the passability prediction sub-module is not activated or the output probability value is less than 60%, the voice playing unit plays the driving of the vehicle management intelligent platform. According to the vehicle positioning, the prompt information of the corresponding road location will be played, and the prompt information will include reasonable vehicle speed and driving lane; otherwise, the decision-making unit will improve the driving plan according to the vehicle management intelligent platform, re-specify the driving speed, and will modify the driving plan. Send it to the voice playback unit for real-time playback; whenever the vehicle reaches the exit of the expressway or is unable to pass in the middle, the wireless communication module returns the information of successful passing or unsafe passing, and the platform records the number of safe passes on all roads on the vehicle's driving path. , the number of unsafe passes, these data are used as training data sets to train the neural network and fuzzy neural network of the vehicle management core module in the vehicle management intelligent platform to improve the accuracy of prediction.

In particular, under certain conditions, the mobile phone can be used as a human-computer interaction interface to collect driver information, and as an in-vehicle wireless communication device to communicate with the vehicle management smart platform, send driver and vehicle information, receive driving plans, and use it as a voice The playing unit broadcasts the route information and reasonable speed value during the driving process.

Compared with the prior art, the beneficial effects of the present invention are:

1. At present, the means of preventing traffic accidents are mainly based on strengthening safety education, and it is expected to improve drivers' awareness of safe driving through safety education. However, this method has poor real-time performance, and the driving ability of different drivers is uneven, and driving experience cannot be achieved. Through the improvement of safety education, the occurrence of traffic accidents is still unavoidable. The present invention can predict the passability of the road ahead in real time during the driver's driving process, and remind the driver to provide the driver's safety awareness. adaptability and real-time.

2. Although the existing assisted driving systems for improving driving safety can guarantee the driving safety to a certain extent, most of them are passively assisted driving, which mainly play a corrective role. It is rare to actively prevent traffic accidents. Before it happens, by predicting the passability of the road and reminding the driver, it can improve the driver's vigilance and fundamentally avoid the occurrence of traffic accidents.

3. The function of the existing assisted driving system for improving driving safety is relatively single, temporarily unable to handle complex traffic environments, and there is no report on the assisted driving system with comprehensive driving assistance functions, and the road passability prediction method of the present invention, Unaffected by complex driving scenarios, it is suitable for various environments.

4. The existing highway sealing and unsealing management methods have a low degree of intelligence, and it is difficult to safely, efficiently and quickly clear the vehicles and manage the driving routes of the vehicles. The road detection system equipped with a wired chassis according to the present invention can detect in time Road conditions, and then determine whether to release, and plan a safe driving route, which can quickly and efficiently dredge vehicles, ensure safety, and make vehicle management more convenient.

Description of drawings:

The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

Fig. 1 is the system composition block diagram of the present invention;

Fig. 2 is the schematic diagram of the road adhesion coefficient estimation method of the present invention;

3 is a schematic diagram of a road curvature estimation method of the present invention;

4 is a schematic diagram of a road gradient estimation method of the present invention;

FIG. 5 is a schematic diagram of the vehicle management method of the present invention.

Detailed ways:

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

Referring to FIG. 1 , a highway road surface detection system with a wire-controlled chassis according to the present invention includes a road surface parameter detection module, a related module on a vehicle, a wireless communication module, a vehicle management intelligent platform, and a voice prompt module; Four modules: vehicle information storage module, voice prompt module, driver information input module and vehicle prediction module. The vehicle-mounted prediction module includes a vehicle-mounted road surface detection sub-module, a passability prediction sub-module, and a data storage sub-module. The vehicle-mounted prediction module is respectively connected with the vehicle information storage module, the voice prompt module and the driver information input module, and the four modules on the vehicle respectively communicate with the vehicle management intelligent platform through the wireless communication module. The intelligent vehicle management platform includes three modules: the core vehicle management module, the historical traffic accident information storage module; the road parameter acquisition module composed of the detection fleet and UAV is connected to the environmental information acquisition module in the vehicle management intelligent platform.

The road parameter detection module is responsible for collecting road parameter information and sending it to the vehicle management intelligent platform, and the environmental information acquisition module receives the information; the driver information input module obtains driver information, the vehicle information storage module collects vehicle information, and the wireless communication module stores the driver information. and vehicle information sent to the vehicle management core module of the vehicle management intelligent platform; the vehicle management intelligent platform judges whether the vehicle can pass safely according to the above information and the information of the historical traffic accident information storage module, and formulates a safe driving route plan for the permitted vehicles. The safe driving plan is sent to the voice prompt module of the vehicle through the wireless communication module; the relevant voice is played by the voice guidance module to guide the vehicle's driving route and speed throughout the process. During the driving process of the vehicle, the on-board detection sub-module in the on-board prediction module is responsible for real-time detection of road parameters. The driver information input module and the vehicle information storage module respectively input the driver information and vehicle information into the passability prediction sub-module. The platform sends the road parameters and safety zone prediction information to the vehicle-mounted prediction module through the wireless communication module. The passability prediction sub-module judges in real time whether the safety level of the current driving safety zone is lowered according to the above input information, and then judges whether it is safe to pass. Store the information on whether the pass is successful or not in the data storage sub-module, and send the information to the vehicle management platform at the exit of the expressway.

Referring to FIG. 2, the method for estimating the road adhesion coefficient of the present invention is to calculate the road adhesion rate according to the tire vertical load and the tire force condition collected by the tire force sensor, estimate the slip rate from the vehicle acceleration and the wheel angular velocity, and finally according to the road adhesion The estimated value ac of the pavement adhesion coefficient is obtained from the rate-slip rate calibration curve.

Referring to FIG. 3 , the road curvature estimation method of the present invention is that the aerial drone captures road images, estimates the road curvature through image recognition, communicates with the probe vehicle wireless communication module, and sends the curvature estimate C1 to the probe vehicle. The GPS receiver obtains the position of the probe car, extracts the line shape of the current road position from the electronic navigation map, and then calculates a road curvature value C2; according to the position of the probe car, find the road curvature design of the corresponding road position in the road design data The road curvature design value C3 corresponding to the current position is obtained; the road curvature design value C3, the road curvature estimated value C1 estimated by the aviation drone, and the road curvature estimated value C2 are weighted and fused to obtain the final road curvature value C. The fusion weight is based on Weather setting: Under severe weather conditions such as rain, snow, fog, and hail, the weight of road curvature value C1 estimated by aviation drones is 0.2, and the weight of road curvature design value C3 is 0.5; under other meteorological conditions, aviation The weight of the road curvature value C1 estimated by the drone is 0.3, the weight of the road curvature value C2 is 0.3, and the weight of the road curvature value C3 is 0.4; when the positioning information cannot be obtained normally, the road curvature value C1 estimated by the aviation drone The weight is 1, the weight of the road curvature value C2 is 0, and the weight of the road curvature value C3 is 0;

Referring to Fig. 4, the road gradient estimation method of the present invention is that an aerial drone captures a road image, estimates the road gradient through image recognition, and obtains the estimated value XA0 of the longitudinal gradient of the road and the estimated value YA0 of the transverse gradient of the road; The road longitudinal gradient information is separated from the original acceleration sensor signal, and then the road longitudinal gradient angle XA1 is obtained; the roll angle of the vehicle body relative to the chassis is estimated through the suspension height sensor information, and finally the road lateral gradient angle YA1 is estimated; using point cloud data Establish an interval grid map in the Cartesian coordinate system, perform plane fitting in the interval to obtain the road surface normal vector, and use the normal vector to calculate the road longitudinal gradient angle XA2 and lateral gradient angle YA2; the longitudinal state of the vehicle based on the vehicle longitudinal dynamics model The observer estimates the longitudinal gradient angle XA3 according to the vehicle torque, rotational speed and acceleration; the vehicle lateral state observer based on the two-degree-of-freedom vehicle kinematics model estimates the road side according to the front wheel rotation angle, the yaw rate of the vehicle body, and the lateral acceleration of the vehicle body Towards the slope angle YA3; the GPS receiver obtains the position of the probe vehicle, searches the design value of the slope angle in the road design data according to the position of the probe car, and obtains the longitudinal gradient angle XA4 and the lateral gradient angle YA4; XA0, XA1, XA2, XA3, XA4 are weighted and fused to obtain the final longitudinal gradient angle XA; the fusion weights of various longitudinal gradient angles are fixed, the longitudinal gradient angle XA1 fusion weight is 0.05, and the longitudinal gradient angle XA1 fusion weight estimated according to the acceleration sensor signal is 0.1, the fusion weight of the longitudinal slope angle XA3 estimated by the longitudinal observer is 0.2, the fusion weight of the slope angle XA2 estimated by the lidar is 0.3, and the fusion weight of the longitudinal slope angle XA4 is 0.5; the above four types of lateral slope angles YA1, YA2, YA3 and YA4 are weighted and fused to obtain the final side slope angle YA; when entering and leaving the side slope, the fusion weight of the side slope angle YA1 calculated based on the information of the suspension height sensor is 0.35. The fusion weight of the lateral slope angle YA3 is 0.15. When on the slope, the fusion weight of the lateral slope angle YA3 estimated by the lateral state observer is 0.35, and the fusion weight of the lateral slope angle YA1 calculated based on the suspension height sensor information is 0.15. The fusion weight of the lateral slope angle YA0 is always 0.05, the fusion weight of the lateral slope angle YA2 estimated from the lidar is always 0.2, and the fusion weight of the lateral slope angle YA4 is always 0.25.

Referring to Fig. 5, the vehicle management method of the present invention is that the vehicle management intelligence platform judges whether the vehicle can pass safely, and formulates a safe driving plan for the vehicle; during the vehicle driving process, the vehicle-mounted prediction module is corrected according to the change of road parameters. Safe driving plan. The vehicle management intelligent platform includes an environmental information acquisition module, a historical traffic accident information storage module, and a core vehicle management module. The environmental information acquisition module acquires visibility, road parameters, time and meteorological information, and the meteorological information includes rain, snow, fog, temperature, and humidity. The historical traffic accident information storage module stores relevant characteristic information of traffic accidents, including the distribution characteristics of the vehicles involved in the accident, the time distribution characteristics of the traffic accidents, the visibility distribution characteristics of the traffic accidents, the meteorological characteristics of the traffic accidents, the location distribution characteristics of the traffic accidents, the density characteristics of the traffic accident vehicles, Traffic accident pavement characteristics and driver characteristics. Among them, the road surface characteristics of traffic accidents refer to the road adhesion coefficient, road curvature and slope, and the characteristics of the driver involved in the accident include the driver's age, gender, driving experience, and occupation.

The core module of vehicle management includes traffic flow distribution sub-unit, safety passing grade evaluation sub-unit and safe driving scheme planning sub-unit. Among them, the traffic flow distribution sub-unit uses the neural network to estimate the type and quantity of vehicles allowed to be released under the current environment and road conditions. The neural network includes an input layer, a hidden layer and an output layer. The input layer has 4 units: time, visibility, weather Vector, road parameters, there are two hidden layers, the first hidden layer is composed of 4 units, the second hidden layer is composed of 2 units, and the output layer has 2 units: the type and number of vehicles that are allowed to be released. The neural network is trained in advance with the distribution characteristics of the vehicles involved in the accident, the time distribution characteristics of the traffic accidents, the location distribution characteristics of the traffic accidents, the visibility distribution characteristics of the traffic accidents, the meteorological characteristics of the traffic accidents, and the density characteristics of the traffic accident vehicles in the historical traffic accident information storage module. Time, visibility, weather, road parameter information is input into the neural network, and the neural network outputs the maximum number of vehicles allowed to pass and the models allowed to pass, and the vehicle intelligence platform selects qualified vehicles to communicate with it according to this.

The safety passing level evaluation sub-unit establishes the vehicle dynamics model according to the vehicle model parameters, tire parameters, braking system parameters, drive system parameters, and steering system parameters stored in the vehicle information storage module. The vehicle dynamics model includes tire model, drive system model , braking system model, vehicle body model, air resistance model. The environmental information acquisition module stores the pavement parameters from the pavement parameter detection module: the pavement adhesion coefficient, the curvature radius of the curve, the longitudinal slope and the lateral slope. According to these pavement parameters, a road model is established. The road types include straight roads and curves with different road adhesion coefficients. , ramps and combinations thereof. The passability, passable speed and transmission gear position of the vehicle under various roads are calculated respectively, and the position of the vehicle can be driven on the road surface, which is input to the fuzzy neural network for predicting the reliability of the driver.

The structure of the fuzzy neural network is a pre-neural network plus a fuzzy neural network, and the pre-neural network is divided into 3 sub-neural networks, namely neural network 1, neural network 2, and neural network 3, and the structure of each sub-neural network is There are 3 layers: input layer, hidden layer, and output layer. The fuzzy neural network structure is divided into 5 layers: input layer, fuzzy layer, fuzzy rule layer, fuzzy decision layer and output layer. The output layer of the three sub-neural networks is fuzzy Part of the input layer of the neural network; train the neural network 1 with the driver's characteristic data, input the basic information of the driver, the basic information of the fellow passengers and the accident characteristics of the driver's traffic accident into the neural network 1, and evaluate the driver's accident probability , in which the accident characteristics of the driver's traffic accident affect the weights of each layer of the neural network, and the neural network 1 finally outputs a value between 0-100%; the neural network 2 is trained with the distribution data of the accident vehicle, and the model parameters of the vehicle are used to train the neural network 2. Input the neural network 2, evaluate the safety level of the vehicle, and output the probability of the vehicle occurrence of a traffic accident, the probability value is between 0-100%; train the neural network 3 with the traffic accident road surface feature data, and input the road parameters of the road that the vehicle can travel in turn. Neural network 3, neural network 3 predicts the probability of traffic accidents caused by these roads, and the probability value is between 0-100%; the input layer of the fuzzy neural network contains 4 units: the driver's probability of causing an accident, the probability of the vehicle's traffic accident , the probability of traffic accidents caused by the road that the vehicle can travel on, the maximum speed limited by the lane, the maximum speed limited by the lane is input multiple times, and the maximum speed limited by one lane at a time, the membership function of the fuzzy layer is determined by the neural network according to historical traffic accidents. Data generation, the fuzzy rules of the fuzzy rule layer are extracted from the knowledge base by the neural network. The neural network adjusts the knowledge base in real time according to the update of traffic accident information, and automatically optimizes the parameters of the fuzzy rules online. There are 4 fuzzy language values: {LPL, MPL , HPL, SPL}, meaning: {low safety pass level, medium safety pass level, high safety pass level, extremely high safety pass level}, the output layer outputs the safety pass level with the highest degree of membership of the road, and input it to Safe driving program planning sub-unit.

The safe driving scheme planning sub-unit uses the safety passing grade output from the safety passing grade to evaluate the safe passing grade of the vehicle's drivable road surface. According to the level of the safety passing grade, priority is given to connecting the road surfaces with the high safety passing grade to form a driving path with the highest comprehensive safety passing grade. ; Roads with lower safety passing grades account for no more than 5% of all road surfaces in the entire driving path, and roads with medium safety passing grades account for less than 10% of all road surfaces in the entire driving path, which are regarded as effective safe paths, and the vehicle is allowed to pass. , enter the waiting area, otherwise it will be regarded as an invalid safety path, the vehicle will not be allowed to pass, and enter the waiting area; the vehicles that can form an effective safety path will be scored according to the comprehensive safety passing grade. The scoring rule is that the safety passing grade is lower and the safety Roads with medium passing grades, high safety passing grades, and extremely high safety passing grades are respectively awarded 25 points, 50 points, 75 points, and 100 points. The total points of each road surface in the accumulated driving path are sorted according to the total, and the traffic flow allows The value releases the corresponding number of vehicles in the top ranking, and transmits the safe route data and the prescribed speed to these vehicles through wireless communication. Vehicles that receive this information can depart in an orderly manner.

During the driving process of the vehicle, if the ratio of the road gradient and road adhesion coefficient calculated by the vehicle-mounted road surface detection sub-module and the difference between the corresponding parameter values measured by the detection vehicle and the corresponding parameter values measured by the detection vehicle exceeds 10%, the road gradient And the road adhesion coefficient is sent to the vehicle management intelligent platform, and the environmental information acquisition module updates the road parameters. And the passability prediction sub-module will be activated to re-evaluate the passability of the current road. If the passability prediction sub-module is not activated or the output safety pass level is not lower than the safety pass level predicted by the vehicle management intelligent platform, the voice playback unit will play the driving plan of the vehicle management intelligent platform, and play the prompt of the corresponding road position according to the vehicle positioning. Information, prompt information includes reasonable speed and driving lane; on the contrary, the decision-making unit will improve the driving plan according to the vehicle management intelligent platform, re-specify the driving speed, and send the revised driving plan to the voice playback unit for real-time playback; whenever the vehicle When reaching the exit of the expressway or if it is impossible to pass, the wireless communication module will return the successful or unsafe passing information, and the platform will record the number of safe passes and unsafe passes on all roads on the vehicle's driving path, and use these data to train the vehicle. The neural network and fuzzy neural network of the core module of vehicle management in the management intelligence platform.

The above discussion is only a preferred embodiment of the present invention, and is for the purpose of explanation and illustration, and not for limiting the present invention itself. The invention is not to be limited to the specific embodiments disclosed herein, but rather is to be determined by the following claims. Furthermore, recitations in the foregoing description in relation to particular embodiments are not to be construed as limitations on the scope of the invention or on the definitions of terms used in the claims. Various other embodiments and various modifications to the disclosed embodiment will be apparent to those skilled in the art. However, all such embodiments, changes and modifications that do not depart from the basic idea of the invention are within the scope of the appended claims.

Claims (8)

1. A highway road surface detection system of a wire-controlled chassis is characterized in that: comprising a road parameter detection module, a wireless communication module, a vehicle management intelligence platform, a vehicle information storage module, a voice prompt module, a driver information input module and an on-board prediction module; wherein, the vehicle information storage module, the voice prompt module, the driver information input module and the vehicle-mounted prediction module are installed on the vehicle, and the vehicle-mounted prediction module includes the vehicle-mounted road surface detection sub-module and the passability prediction sub-module; the wireless communication module includes the vehicle-mounted wireless The transceiver unit and the wireless transceiver unit of the intelligent platform; the intelligent vehicle management platform includes an environmental information acquisition module, a historical traffic accident information storage module, and a core vehicle management module; the road parameter detection module is responsible for collecting road parameter information and sending it to the vehicle management intelligent platform; the driver The information input module obtains driver information, the vehicle information storage module collects vehicle information, and the wireless communication module sends the driver information and vehicle information to the vehicle management intelligent platform; the vehicle management core module of the vehicle management intelligent platform judges whether the vehicle can be released according to the above information. And make a safe driving plan for the vehicles that are allowed to be released, and send the safe driving plan to the voice prompt module of the vehicle through the wireless communication module; the voice guidance module plays the relevant voice to guide the vehicle's driving route and speed throughout the whole process; during the driving process of the vehicle, the vehicle predicts The on-board road detection sub-module in the module is responsible for real-time detection of road parameters. The driver information input module and vehicle information storage module respectively input driver information and vehicle information into the passability prediction sub-module. The prediction information is sent to the vehicle-mounted prediction module through the wireless communication module, and the passability prediction sub-module judges in real time whether the safety level of the current driving safety zone is lowered according to the above input information, and then judges whether it is safe to pass, and stores the information on the success of the passage. to the data storage sub-module, and send the information to the vehicle management intelligence platform at the highway exit. 2. The highway road surface detection system of a wire-controlled chassis according to claim 1, characterized in that: the road surface parameter detection module comprises a detection vehicle and an aviation drone; wherein, the detection vehicle is based on a wire-controlled chassis The wheelbase, wheelbase and body height can be changed within a certain range. During the detection process, the area with the worst road adhesion conditions in front of the road is identified by aerial drone image recognition, and the wheelbase, vehicle Adjust the distance accordingly to adapt to the area with the worst road conditions and detect the worst road adhesion coefficient; the tires are divided into two types of detection vehicles using passenger tires and commercial tires, each equipped with 8 wheels, four types of tires, front and rear tires. Two tires on the left and right are arranged side by side, and the tire pressure can be adjusted dynamically, so that the tires of the detection vehicle can be equipped and the tire pressure can be adjusted according to the tire type and load of the vehicle to be passed; One for each lane, driving at the specified speed of the lane. In the first detection, the detection vehicles are sent from the starting points of the two driving directions of a highway at the same time, that is, the two teams of detection vehicles are driving towards each other; the detection vehicles of the two teams will detect The road adhesion coefficient is uploaded to the vehicle management intelligent platform, and the vehicle management intelligent platform compares the adhesion conditions of the lanes in the two directions. If the road adhesion coefficients at the same position are similar, the second time the probe vehicle is dispatched, only one direction is required. The detection vehicle can collect road parameters. 3. The highway road surface detection system of a wire-controlled chassis according to claim 1, characterized in that: the method for acquiring road parameters by the detection vehicle in the road surface parameter detection module is that the detection vehicle has an information acquisition unit, Road curvature calculation unit, road adhesion coefficient calculation unit, road gradient calculation unit; wherein, the information acquisition unit obtains suspension height sensor, inertial measurement unit, lidar, tire force sensor, vehicle acceleration sensor and wheel angular velocity sensor, global positioning system respectively The signal of the receiver and the electronic navigation map, the aerial drone captures the road image 500 meters in front of the vehicle, estimates the road slope and curvature through image recognition, and communicates with the wireless communication module of the rover, and converts the road longitudinal slope estimated value XA0, road The estimated lateral slope value YA0 and the estimated value of curvature C1 are sent to the probe vehicle; the above information collected by the information acquisition unit is used by the road curvature calculation unit, the road adhesion coefficient calculation unit, and the road gradient calculation unit; the road adhesion coefficient calculation unit is based on the tire vertical load and The tire force collected by the tire force sensor is used to calculate the road adhesion rate, and the slip rate is estimated from the vehicle acceleration and wheel angular velocity. Finally, the road adhesion coefficient estimated value ac is obtained according to the road adhesion rate-slip rate calibration curve, which is calculated every 5 meters. The estimated value of the adhesion coefficient ac; the position of the probe car is obtained by the GPS receiver, and the road curvature calculation unit extracts the line shape of the current road position from the electronic navigation map, and then calculates a road curvature value C2; Find the road curvature design value of the corresponding road position in the data, and obtain the road curvature design value C3 corresponding to the current position. The final road curvature value C, the fusion weight is set according to the weather conditions: under severe weather conditions such as rain, snow, fog, and hail, the weight of the road curvature value C1 estimated by the aviation drone is 0.2, and the weight of the road curvature design value C3 is 0.5; under other meteorological conditions, the weight of the road curvature value C1 estimated by the aviation drone is 0.3, the weight of the road curvature value C2 is 0.3, and the weight of the road curvature value C3 is 0.4; When the positioning information cannot be obtained normally, The weight of the road curvature value C1 estimated by the aviation drone is 1, the weight of the road curvature value C2 is 0, and the weight of the road curvature value C3 is 0; The point cloud data is input into the gradient calculation subunit, and the gradient calculation subunit uses the least squares method to separate the road longitudinal gradient information from the original acceleration sensor signal, and then obtains the road longitudinal gradient angle XA1; the vehicle body relative to the chassis is estimated by the suspension height sensor information and finally estimate the road lateral slope angle YA1; use the point cloud data to establish an interval grid map in the Cartesian coordinate system, perform plane fitting within the interval to obtain the road surface normal vector, and use the normal vector to calculate the road longitudinal slope angle XA2 and lateral slope angle YA2; at the same time, the slope The vehicle longitudinal state observer based on the vehicle longitudinal dynamics model in the calculation subunit estimates the longitudinal slope angle XA3 according to the vehicle torque, rotational speed and acceleration; The yaw rate of the vehicle body and the lateral acceleration of the vehicle body are used to estimate the lateral slope angle YA3 of the road; the GPS receiver obtains the position of the probe car, and the design value of the slope angle is searched in the road design data according to the position of the probe car to obtain the longitudinal slope angle XA4 and lateral slope angle YA4; the above five types of longitudinal slope angles XA0, XA1, XA2, XA3, XA4 are weighted and fused to obtain the final longitudinal slope angle XA; the fusion weight of various longitudinal slope angles is fixed, and the fusion weight of longitudinal slope angle XA1 is 0.05 , the fusion weight of the longitudinal slope angle XA1 estimated according to the acceleration sensor signal is 0.1, the fusion weight of the longitudinal slope angle XA3 estimated by the longitudinal observer is 0.2, the fusion weight of the slope angle XA2 estimated according to the lidar is 0.3, and the fusion weight of the longitudinal slope angle XA4 is 0.5; the above four types of lateral slope angles YA1, YA2, YA3, and YA4 are weighted and fused to obtain the final lateral slope angle YA; when entering and leaving the lateral slope, the lateral slope calculated based on the information of the suspension height sensor The angle YA1 has a fusion weight of 0.35, the lateral slope angle YA3 estimated by the lateral state observer has a fusion weight of 0.15, and when on a slope, the lateral slope angle YA3 estimated by the lateral state observer has a fusion weight of 0.35, based on the suspension height. The fusion weight of the lateral slope angle YA1 calculated from the sensor information is 0.15, the fusion weight of the lateral slope angle YA0 is always 0.05, the fusion weight of the lateral slope angle YA2 estimated from the lidar is always 0.2, and the fusion weight of the lateral slope angle YA4 is always 0.2. The weight is always 0.25. 4. The highway road surface detection system of a wire-controlled chassis according to claim 1, wherein the driver information input module comprises a human-machine interface and a data storage device, and through the human-machine interface, to In the form of the questionnaire, the driver inputs the basic information of the driver, the basic information of the accompanying passengers and the accident characteristics of the driver's previous traffic accidents; the basic information of the driver includes age, gender, ID number, driving experience, average daily driving time, physical Health status, mental status and occupation; driving age is input with natural numbers 0, 1, 2..., unit is year; average daily driving time allows input of an integer between 0-24, unit is hour; there are three options for physical health status: Health, sub-health, disease states, give corresponding notes below the options for the driver to understand, "sick state" refers to one of the symptoms of cold, headache, fever, and other parts of the body in "sub-health state" pain or feel Unpleasant but difficult to describe, "health status" refers to normal body, no disease and pain; mental status provides three options: good, poor, very poor, also give corresponding notes under the options for the driver to understand, "very poor" ” means feeling very drowsy and having difficulty concentrating, “poor” means feeling drowsy and somewhat inattentive, and “good” means not drowsy and able to concentrate, which is input after the driver’s self-assessment; Occupation options provide two categories: professional drivers and non-professional drivers; accident characteristics include the number of accidents, accident form, accident severity, accident cause, type of vehicle involved, and accident time. Among them, accident forms include rear-end collision, scratch, and collision fix The severity of the accident is divided into no casualty accident, injury accident, and fatal accident. The cause of the accident is driving in the same lane, not keeping a safe distance from the vehicle in front as required, improper operation, not using lights as required or not following the required speed in low visibility. Driving, illegal lane change, improper braking, traffic signal violation, fatigue driving, illegal driving on the road, the types of vehicles involved include heavy trucks, medium trucks, light trucks, minivans, large passenger vehicles, medium passenger vehicles, light passenger vehicles, small passenger vehicles and passenger vehicles When using a car, the accident time is specific to the date and time; the data storage device stores the above information. 5 . The highway road surface detection system of a chassis-by-wire according to claim 1 , wherein the vehicle information storage module stores vehicle parameters, tire information, braking system information, drive system information, and steering system information. 6 . Information, the age of the vehicle, the mileage, the accident characteristics of the vehicle's historical traffic accidents, and the model parameters include vehicle type, vehicle length, width and height, minimum turning radius, vehicle curb weight, maximum driving force, total vehicle mass, axle Distance, height of center of mass, distance from center of mass to rear axle, windward area, wheelbase, wheelbase, minimum ground clearance, approach angle, departure angle, maximum speed, maximum output torque, number of wheels and driving wheels, automatic driving level ;Tire information refers to tire type and tire radius; Brake system information includes brake type and size parameters; Drive system information includes power type and related size parameters; Steering system information includes steering system type and related size parameters; Vehicle types include commercial vehicles , passenger vehicles, and commercial vehicles are subdivided into heavy trucks, medium trucks, light trucks, mini trucks, large passenger cars, medium passenger vehicles, light passenger vehicles, and small passenger vehicles; The time from the start of recording at one time to the current moment is used as the vehicle's service life; the accident characteristics of the vehicle's historical traffic accidents include the number of accidents, accident form, damage and accident time. Hit the fixed object, the damage includes scratched parts, damaged parts, repaired or replaced parts, and the accident time is recorded in the year, month, and day. 6. The highway road surface detection system of a wire-controlled chassis according to claim 1, wherein the environmental information acquisition module acquires visibility, road parameters, time and weather information, and the weather information includes rain, snow, Fog, temperature, and humidity; the historical traffic accident information storage module stores characteristic information about traffic accidents, including the distribution characteristics of the vehicles involved, the time distribution characteristics of traffic accidents, the visibility distribution characteristics of traffic accidents, the meteorological characteristics of traffic accidents, the location distribution characteristics of traffic accidents, Traffic accident vehicle density characteristics, traffic accident road surface characteristics, and accident driver characteristics; among which, traffic accident road surface characteristics refer to road adhesion coefficient, road curvature and slope, and accident driver characteristics include the age, gender, driving experience, and occupation of the accident driver; The core module of vehicle management includes a traffic flow distribution sub-unit, a safe passing grade evaluation sub-unit and a safe driving scheme planning sub-unit; among them, the traffic flow distribution sub-unit determines the number and type of vehicles to be released, and uses a neural network to estimate the current environment and road conditions. The model and number of vehicles allowed to be released under the neural network include input layer, hidden layer and output layer. The input layer has 4 units: time, visibility, weather vector, and road parameters. There are two hidden layers. The first hidden layer consists of It consists of 4 units, the second hidden layer consists of 2 units, and the output layer has 2 units: the type and number of vehicles allowed to be released, the distribution characteristics of the vehicles involved in the accident in the historical traffic accident information storage module, and the time distribution of traffic accidents in advance. characteristics, traffic accident location distribution characteristics, traffic accident visibility distribution characteristics, traffic accident meteorological characteristics and traffic accident vehicle density characteristics to train this neural network, input time, visibility, weather, road parameter information into the neural network, and the neural network outputs allowable traffic. The maximum number of vehicles and the types of vehicles allowed; The safety passing grade evaluation sub-unit evaluates the passability of the vehicle on a specific road and the driving reliability of the driver on a specific road respectively, and evaluates the safety passing grade by combining the results of these two aspects. Establish vehicle dynamics model and road model for calculation. The dynamics model includes tire model, drive system model, braking system model, vehicle body model, and air resistance model. According to the vehicle model parameters, tire parameters, brake parameters stored in the vehicle information storage module System parameters, drive system parameters, and steering system parameters are used to establish the dynamic model of the vehicle. The road model includes 4 parameters: road adhesion coefficient, curve radius of curvature, longitudinal slope and lateral slope. The data comes from the road parameter detection module; road types include The straights, curves, ramps and their combinations of different road adhesion coefficients are used to calculate the passability of the vehicle under various roads, the passable speed and transmission gear position, and obtain the position on the road where the vehicle can travel; The driving reliability of the driver on a specific road uses the fuzzy neural network to predict the probability of the driver driving on the drivable road. The structure of the fuzzy neural network is a pre-neural network plus a fuzzy neural network, and the pre-neural network is divided into 3 sub-systems The neural network is neural network 1, neural network 2, and neural network 3. The structure of each sub-neural network is 3 layers: input layer, hidden layer, and output layer. The fuzzy neural network structure is divided into 5 layers: input layer. , fuzzy layer, fuzzy rule layer, fuzzy decision layer and output layer, the output layer of the three sub-neural networks is part of the input layer of the fuzzy neural network. The basic information of passengers and the accident characteristics of the driver's traffic accidents are input into the neural network 1 to evaluate the driver's accident probability. The accident characteristics of the driver's previous traffic accidents affect the weights of each layer of the neural network, and the neural network 1 finally Output the value between 0-100%; train the neural network 2 with the distribution data of the vehicles involved in the accident, input the model parameters of the vehicle into the neural network 2, evaluate the safety level of the vehicle, and output the probability of the vehicle having a traffic accident, the probability value is 0-100 %; train the neural network 3 with the traffic accident pavement feature data, and input the road parameters of the vehicle's drivable road surface into the neural network 3 in turn, and the neural network 3 predicts the probability of traffic accidents caused by these roads, and the probability value is between 0-100% ;The input layer of the fuzzy neural network contains 4 units: the driver's probability of causing an accident, the probability of the vehicle's traffic accident, the probability of a traffic accident caused by the vehicle's drivable road surface, the maximum speed of the lane limit, and the maximum speed of the lane limit is input multiple times. Vehicle speed, input the maximum speed limited by one lane at a time, the membership function of the fuzzy layer is generated by the neural network according to the historical traffic accident data, the fuzzy rules of the fuzzy rule layer are extracted from the knowledge base by the neural network, and the neural network is based on the traffic accident information. Update the real-time adjustment knowledge base, and automatically optimize the parameters of the fuzzy rules online. There are 4 fuzzy language values: {LPL, MPL, HPL, SPL}, meaning: {The safety pass level is low, the safety pass level is medium, the safety pass level is relatively high High, the safety pass level is extremely high}, the output layer outputs the safety pass level with the highest degree of membership of the road; The safe driving scheme planning sub-unit uses the safety passing grade output from the safety passing grade to evaluate the safe passing grade of the vehicle's drivable road surface. According to the level of the safety passing grade, priority is given to connecting the road surfaces with the high safety passing grade to form a driving path with the highest comprehensive safety passing grade. ; Roads with low safety passing grades account for no more than 5% of all roads in the entire driving path, and roads with moderate safety passing grades account for less than 10% of all road surfaces in the entire driving path. They are considered effective safe paths, and vehicles are allowed to pass. Enter the waiting area, otherwise it will be regarded as an invalid safety path, and vehicles will not be allowed to pass, and enter the waiting area; the vehicles that can form an effective safety path will be scored according to the comprehensive safety passing grade. The scoring rule is that the safety passing grade is lower and the safety passing grade is lower Roads with medium, high safety passing grades, and extremely high safety passing grades are respectively awarded 25 points, 50 points, 75 points, and 100 points. The total score of each road surface in the accumulated driving path is sorted according to the total level, and released according to the allowable value of traffic flow. The corresponding number of vehicles ranked at the top, and the safe route data and the prescribed speed are sent to these vehicles through wireless communication. 7. The highway road surface detection system of a wire-controlled chassis according to claim 1, characterized in that: the vehicle-mounted road surface detection sub-module will detect the current road parameters, and the passability prediction sub-module is based on the detected Pavement parameter conditions, refer to the safety level evaluated by the platform, and re-evaluate the passability of the current road surface; the vehicle road surface detection sub-module includes an information acquisition sub-unit, a road adhesion coefficient calculation sub-unit, and a road gradient calculation sub-unit; among them, the information acquisition unit Obtain the signals of suspension height sensor, inertial measurement unit, lidar, tire force sensor, vehicle acceleration sensor and wheel angular velocity sensor respectively, which are used by the road adhesion coefficient calculation subunit and the road gradient calculation subunit; The road adhesion coefficient calculation unit calculates the road adhesion rate according to the tire vertical load and the tire force collected by the tire force sensor, estimates the slip rate from the vehicle acceleration and wheel angular velocity, and finally obtains the road adhesion coefficient according to the road adhesion rate-slip rate calibration curve. Estimated value ac', detect and calculate an estimated value of adhesion coefficient ac' at the starting position of the road surface for each specific safety level; The information acquisition subunit inputs the acceleration, wheel torque and rotational speed signals, and the point cloud data generated by the lidar into the gradient calculation subunit. The gradient calculation subunit uses the least squares method to separate the road longitudinal gradient information from the original acceleration sensor signal, and then obtains the road Longitudinal slope angle XA1'; the roll angle of the vehicle body relative to the chassis is estimated through the information of the suspension height sensor, and finally the road lateral slope angle YA1' is estimated; the point cloud data is used to establish the interval grid map in the Cartesian coordinate system, The road surface normal vector is obtained by plane fitting within the interval, and the normal vector is used to calculate the road longitudinal gradient angle XA2' and the lateral gradient angle YA2'; at the same time, the vehicle longitudinal state observer based on the vehicle longitudinal dynamics model in the gradient calculation subunit is based on the The vehicle torque, rotational speed and acceleration estimate the longitudinal gradient angle XA3'; the vehicle lateral state observer based on the two-degree-of-freedom vehicle kinematics model estimates the road lateral gradient according to the front wheel rotation angle, the vehicle body yaw rate, and the vehicle body lateral acceleration Angle YA3'; the above three types of longitudinal gradient angles XA1', XA2', XA3' are weighted and fused to obtain the final longitudinal gradient angle XA'; the fusion weights of various longitudinal gradient angles are fixed based on the acceleration sensor signals. The estimated gradient angle XA1' fusion weight is 0.2, the fusion weight of the longitudinal slope angle XA3' estimated by the longitudinal observer is 0.3, and the fusion weight of the slope angle XA2' estimated by the lidar is always 0.5; the above three types of lateral slope angles YA1', YA2', YA3' are weighted Fusion to get the final lateral slope angle YA'; when entering and leaving the lateral slope, the lateral slope angle YA1' calculated based on the information of the suspension height sensor is fused with a weight of 0.35, and the lateral slope angle estimated by the lateral state observer is 0.35. The fusion weight of the slope angle YA3' is 0.15. When on the slope, the fusion weight of the lateral slope angle YA3' estimated by the lateral state observer is 0.35, and the fusion weight of the lateral slope angle calculated based on the suspension height sensor information is 0.15. According to The YA2 fusion weight of the lateral slope angle estimated by the lidar is 0.5; If the ratio of the difference between the road gradient and road adhesion coefficient calculated by the vehicle-mounted road surface detection sub-module and the corresponding parameter value measured by the rover vehicle and the corresponding parameter value measured by the rover vehicle exceeds 10%, the passability prediction sub-module will be activated. Re-evaluate the passability of the current road; the passability prediction sub-module uses the fuzzy neural network to predict the road passability of the model. The fuzzy neural network prediction model structure is a pre-neural network plus a fuzzy neural network, and the pre-neural network is divided into 3 The sub-neural networks are neural network 1, neural network 2, and neural network 3. The structure of each sub-neural network is 3 layers: input layer, hidden layer, and output layer. The fuzzy neural network structure is divided into 5 layers: input layer, fuzzy layer, fuzzy rule layer, fuzzy decision layer and output layer, the output layer of the three sub-neural networks is a part of the input layer of the fuzzy neural network; the neural network 1 is trained with the characteristic data of the driver involved in the accident, and the basic information of the driver and the The accident characteristics of the driver's previous traffic accident are input into the neural network 1 to evaluate the driver's accident probability, and finally output the value between 0-100%; the neural network 2 is trained with the distribution data of the accident vehicle, and the parameters of the vehicle type are input into the neural network. Network 2, evaluates the safety level of the vehicle, and outputs the probability of a traffic accident on the vehicle, with the probability value between 0-100%; trains the neural network 3 with the road surface characteristic data of the traffic accident, and inputs the collected road parameters into the neural network 3 , predict the probability of a road accident causing a traffic accident, the probability value is between 0-100%; the input layer of the fuzzy neural network contains 5 units: the driver's probability of causing an accident, the probability of the vehicle causing a traffic accident, and the road causing traffic accidents. The probability of the traffic accident caused by the environment, the speed of the vehicle, the membership function of the fuzzy layer is generated by the neural network based on the historical traffic accident data, the fuzzy rules of the fuzzy rule layer are extracted from the knowledge base by the neural network, and the neural network is based on the traffic accident information. The updated knowledge base is adjusted in real time, and the parameters of fuzzy rules are automatically optimized online. There are 4 fuzzy language values: {LPL, MPL, HPL, SPL}, meaning: {low safety pass level, medium safety pass level, safety pass level higher, the safety pass level is extremely high}, the output layer outputs the safety pass level with the highest degree of membership of the road. 8. The highway road surface detection system of a wire-controlled chassis according to claim 1, characterized in that: the voice prompting module comprises a voice playing unit and a decision-making unit; The safety pass level of the module not activated or output is not lower than the safety pass level predicted by the vehicle management intelligent platform. The voice playback unit plays the driving plan of the vehicle management intelligent platform, and plays the prompt information of the corresponding road position according to the vehicle positioning. The prompt information includes reasonable Vehicle speed and driving lane; on the contrary, the decision-making unit will improve the driving plan according to the vehicle management intelligent platform, re-specify the driving speed, and send the revised driving plan to the voice playback unit for real-time playback; whenever the vehicle reaches the highway exit or midway In the event of impassability, the wireless communication module returns information of successful or unsafe passage, and the platform records the number of safe passages and unsafe passages on all roads on the vehicle's driving path, and uses these data to train vehicle management in the vehicle management smart platform. The core module of neural network and fuzzy neural network.

Images