CN115077537A - High-precision map perception container design method and device, storage medium and terminal - Google Patents

Abstract

A multi-vehicle joint perception oriented high-precision map perception container design method and device, a storage medium and a terminal are provided, and the method comprises the following steps: acquiring perception target indication information, wherein the perception target indication information is used for indicating environmental state information required to be perceived by an automatic driving automobile; obtaining a high-precision map, establishing a map perception container, and performing feature extraction on at least a part of perception targets in the map to obtain sensor information, so as to generate signals with the same format as that of a common sensor according to map original data. In addition, considering that some information in the high-precision map data has the confidence degree far beyond the vehicle-mounted sensor, the information is directly superposed on the sensing result. Finally, carrying out fusion perception on the multi-vehicle combined perception information under a unified map space-time reference to determine a perception result of the automatic driving vehicle; and the positioning error of the map is within a preset error range. The invention can effectively eliminate the perception blind area and improve the perception precision.

Description

Design method and device, storage medium and terminal for high-precision map perception container

technical field

The invention relates to the field of perception of autonomous vehicles, in particular to a design method and device, a storage medium and a terminal for a high-precision map perception container oriented to multi-vehicle joint perception.

Background technique

Bicycle environment perception has bottlenecks such as difficulty in eliminating blind spots and low perception accuracy. V2X (Vehicle ToEverything) Internet of Vehicles can realize the communication between vehicles and vehicles, vehicles and people, vehicles and roadside, and vehicles and cloud platforms, realize the sharing of information, and make autonomous driving bid farewell to the era of bicycle intelligence.

In recent years, with the development of Internet of Vehicles technology, the range of information that can be obtained by bicycles has greatly increased. By integrating the perception information of other vehicles in the Internet of Vehicles, it can effectively eliminate vehicle perception occlusion and super-view blind spots, and improve the perception ability of bicycles; in addition, the prior information of the road environment can also be obtained through the electronic map, which further improves the integration of external perception resources. Bicycle perception provides the technical foundation.

However, in the existing autonomous vehicle perception technology based on the Internet of Vehicles technology, the local coordinate system of the main vehicle is used as the information fusion benchmark, and it depends on the overlapping degree of vision between multiple vehicles. Constraints, there are blind spots in the perception of bicycles, and it is difficult to eliminate them through the visual field data of other connected vehicles.

In the prior art, there are studies on multi-vehicle joint perception based on the Global Navigation Satellite System (GNSS) benchmark. However, in complex scenarios, especially small-scale joint perception scenarios, the GNSS coordinate system is unstable, and there are often edges along the way. The overall drift in a certain direction causes the accuracy of the fusion perception of the network to decrease.

There is an urgent need for a high-precision map perception container design method for multi-vehicle joint perception, which can provide a unified benchmark for information fusion of multi-vehicle joint perception. The benchmark does not drift with the degradation of GNSS signals, and can provide multi-vehicle joint perception results. Stable spatial reference, effectively eliminating blind spots and improving perception accuracy.

SUMMARY OF THE INVENTION

The technical problem solved by the present invention is to provide a high-precision map sensing container design method and device, storage medium and terminal for multi-vehicle joint sensing, which can effectively eliminate sensing blind spots and improve sensing accuracy.

To solve the above technical problem, an embodiment of the present invention provides a high-precision map perception container design method for multi-vehicle joint perception, including the following steps: acquiring perception target indication information, where the perception target indication information is used to indicate the automatic driving. The environment state information that the car needs to perceive; obtain a high-precision map, establish a map perception container, and perform feature extraction on at least a part of the perception targets in the map to obtain sensorized information; extract the high-precision map data in addition to the above-mentioned sensorized information The non-sensorized information is superimposed with ordinary sensor perception information and map sensor perception information, and fusion perception is performed under the unified map space-time reference, and the environment perception result of the autonomous vehicle is obtained by solving the problem through state estimation; The map is the map with the positioning error within the preset error range.

Optionally, the at least a part of the perception targets include connected vehicles and other obstacles other than the connected vehicles; wherein the sensorized information includes connected vehicle detection information for detecting connected vehicles and information on connected vehicles other than the connected vehicles. Non-connected vehicle detection information that detects obstacles other than the vehicle.

Optionally, the non-networked vehicle detection information includes one or more of the following: dynamic obstacle information for detecting dynamic obstacles other than the networked vehicle; detecting static obstacles on the map map detection information.

Optionally, using a map sensor to perform feature extraction on at least a part of the sensing targets in the map to obtain sensorized information includes: determining the position information of the sensing target to be sensed according to the sensing target indication information; The accuracy of the map is calculated, and the covariance of the position information of the at least a part of the perceptual target is calculated as the sensorized information.

Optionally, the non-sensorized information other than the above-mentioned sensorized information in the extracted high-precision map data is superimposed with the common sensor perception information and the map sensor perception information, and fusion perception is performed under a unified map space-time reference, through The state estimation solution problem to obtain the environmental perception result of the autonomous vehicle includes: acquiring sensor information of one or more sensors for detecting the at least a part of the perception target; Fusion to obtain fusion information of each sensing target in the at least a part of the sensing targets; according to the fusion information, determine the observation set of each sensing target in the at least a part of the sensing targets; using a maximum likelihood estimation algorithm, according to the A set of observation quantities of each sensing target in at least a part of the sensing targets, and determining the state quantity estimation of each sensing target in the at least part of the sensing targets.

Optionally, the non-sensorized information other than the above-mentioned sensorized information in the extracted high-precision map data is superimposed with the common sensor perception information and the map sensor perception information, and fusion perception is performed under a unified map space-time reference, through Solving the problem of state estimation to obtain the environmental perception result of the self-driving car further includes: superimposing the state quantity estimation of each perception target in the at least a part of the perception targets and the non-sensorized information to obtain the perception result of the self-driving car.

Optionally, the following formula is used to determine the state quantity estimation of each sensing target in the at least a part of the sensing targets according to the observation quantity set of each sensing target in the at least a part of the sensing targets:

用于表示极大似然意义下驾驶环境空间状态参数的估计结果，X^D用于表示驾驶环境空间中一定时间深度内的观测量集合，Z^p用于表示智能汽车可获取的异步异构的感知空间中的测量集合，P(Z^p|X^D)是在给定状态参数X^D下观测量集合Z^p的条件概率。in,

It is used to represent the estimation results of the state parameters of the driving environment space in the sense of maximum likelihood, X ^D is used to represent the set of observations within a certain time depth in the driving environment space, and Z ^p is used to represent the asynchronous and heterogeneous data available to smart cars. The set of measurements in perceptual space, P(Z ^p |X ^D ) is the conditional probability of observing the set of measurements Z ^p given the state parameter X ^D .

In order to solve the above technical problem, an embodiment of the present invention provides a high-precision map perception container design device for multi-vehicle joint perception, including: an information acquisition module for acquiring perception target indication information, the perception target indication information is used to indicate The self-driving car needs to perceive environmental state information; a sensorized information determination module is used to obtain a high-precision map, establish a map perception container, and perform feature extraction on at least a part of the perception targets in the map to obtain sensory information; perception; The result solving module is used to extract the non-sensorized information other than the above-mentioned sensorized information in the high-precision map data, superimpose it with the common sensor perception information and map sensor perception information, and perform fusion perception under the unified map space-time reference. The state estimation solution problem obtains the environment perception result of the autonomous vehicle; wherein, the high-precision map is a map with a positioning error within a preset error range.

In order to solve the above technical problem, an embodiment of the present invention provides a storage medium on which a computer program is stored, and the computer program executes the steps of the above-mentioned multi-vehicle joint perception-oriented high-precision map perception container design method when the computer program is run by a processor.

To solve the above technical problem, an embodiment of the present invention provides a terminal, including a memory and a processor, the memory stores a computer program that can run on the processor, and is characterized in that the processor runs the The computer program executes the steps of the above-mentioned multi-vehicle joint perception-oriented high-precision map-aware container design method.

Compared with the prior art, the technical solutions of the embodiments of the present invention have the following beneficial effects:

In the embodiment of the present invention, by acquiring sensing target indication information, the sensing target indication information is used to indicate the environmental state information that the self-driving car needs to perceive; acquiring a high-precision map, establishing a map perception container in the map Feature extraction is performed on at least a portion of the perceptual objects to obtain sensorized information, thereby generating a signal in the same format as a common sensor based on the map raw data. In addition, considering that some information in the high-precision map data has far more confidence than the on-board sensors, it is directly superimposed on the perception results. Finally, the multi-vehicle joint perception information (including ordinary sensor perception information, map sensor perception information, and other non-sensorized data in high-precision maps) is fused under the unified map space-time reference to determine the perception result of the autonomous vehicle; wherein , the positioning error of the map is within a preset error range. The present invention can effectively eliminate the perception blind area and improve the perception precision.

Further, acquiring sensor information of one or more sensors for detecting the at least a part of the perceptual targets, and then obtaining fusion information and an observation set of each perceptual target in the at least a part of the perceptual targets, and then determining the at least one part of the perceptual targets. The state quantity estimation of each sensing target can realize the determination of the sensing result based on the information fusion of the map sensor and the sensor, and further improve the sensing accuracy.

Further, non-sensorized information other than the sensorized information may also be extracted from the map, and then the state quantity estimates of each of the at least a part of the perceptual objects are superimposed on the non-sensorized information to obtain The perception result of the self-driving car is obtained, so that other information of the high-precision map can be further added on the basis of the detection of various sensors, so as to further improve the overall effect of perception.

Description of drawings

1 is a flow chart of a method for designing a high-precision map-aware container for multi-vehicle joint perception in an embodiment of the present invention;

Fig. 2 is a flow chart of a specific implementation of step S13 in Fig. 1;

3 is a schematic structural diagram of a high-precision map perception container design device for multi-vehicle joint perception in an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a sensing system based on a multi-vehicle joint sensing-oriented high-precision map sensing container design method according to an embodiment of the present invention.

Detailed ways

As mentioned above, in the prior art, by integrating the perception information of other vehicles in the Internet of Vehicles, it is possible to effectively eliminate vehicle perception occlusion and over-the-field blind spots, and improve the perception capability of a bicycle. However, in the existing autonomous vehicle perception technology based on vehicle network technology, the local coordinate system of the main vehicle is often used as the information fusion benchmark, and it depends on the overlapping degree of vision between multiple vehicles, which fails to form an effective vehicle pose. Constraints, the joint perception accuracy based on mass-produced vehicle sensors is difficult to meet the needs of intelligent networked vehicles, and the consistency of fusion results between multiple sensors is insufficient.

The inventors of the present invention have found through research that, in the existing research scenario of multi-vehicle joint perception based on GNSS benchmarks, due to the instability of the GNSS coordinate system, there is often an overall drift along a certain direction, which results in the accuracy of network-connected fusion perception. decline. The high-precision map data can provide high-precision absolute position constraints for vehicles. At the same time, high-precision map data contains precise road attributes and topology information that is difficult for sensors to perceive.

In the embodiment of the present invention, the sensing target indication information is obtained, and the sensing target indication information is used to indicate the environmental state information that the self-driving car needs to perceive; a high-precision map is obtained, and a map sensing container is established in the map for at least Feature extraction is performed on a part of the perceptual targets to obtain sensorized information, thereby generating signals in the same format as ordinary sensors based on the original map data. In addition, considering that some information in the high-precision map data has far more confidence than the on-board sensors, it is directly superimposed on the perception results. Finally, the multi-vehicle joint perception information (including ordinary sensor perception information, map sensor perception information, and other non-sensorized data in high-precision maps) is fused under the unified map space-time reference to determine the perception result of the autonomous vehicle; wherein , the positioning error of the map is within a preset error range. The present invention can effectively eliminate the perception blind area and improve the perception precision.

In order to make the above objects, features and beneficial effects of the present invention more clearly understood, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

Referring to FIG. 1 , FIG. 1 is a flowchart of a method for designing a high-precision map perception container for multi-vehicle joint perception in an embodiment of the present invention. The high-precision map perception container design method for multi-vehicle joint perception may include steps S11 to S13:

Step S11: acquiring sensing target indication information, where the sensing target indication information is used to indicate the environmental state information that the self-driving car needs to perceive;

Step S12: acquiring a high-precision map, establishing a map perception container, and performing feature extraction on at least a part of the perception targets in the map to obtain sensorized information;

Step S13: Extract the non-sensorized information in the high-precision map data except the above-mentioned sensory information, superimpose with the common sensor perception information and map sensor perception information, perform fusion perception under the unified map space-time reference, and solve the problem through state estimation The problem gets the results of environment perception for self-driving cars.

The high-precision map is a map with a positioning error within a preset error range.

In the specific implementation of step S11 , the sensing target indication information may be determined in a preset manner, that is, a preset sensing requirement may be determined, or a user input manner may be used, so that the user's sensing requirement may be determined.

The sensing target indication information is used to indicate the environmental status information that the autonomous driving vehicle needs to perceive, and the environmental status information may include the spatial range that the autonomous driving vehicle needs to perceive, and may also belong to the environmental status information that the autonomous driving vehicle needs to perceive. One or more types of sensing targets, such as buildings, vehicles, people, etc., that need to be sensed for autonomous vehicles.

In the specific implementation of step S12, a high-precision map may be set.

Specifically, the positioning error of the map is within a preset error range. It can be understood that, the smaller the error value in the preset error range is, the higher the accuracy requirement of the map is.

As a non-limiting example, the upper limit of the error value in the preset error range may be set to be selected from: 10 cm to 50 cm. For example, the upper limit of the error value in the preset error range may be set to 20 cm, that is, the positioning error of the map is within 20 cm.

In a specific implementation, a map-aware container may also be established.

Specifically, the map perception container is based on the high-precision map, and is a theoretical model for solving the problem of joint perception of intelligent vehicles. The driving environment space is constructed under the benchmark, and the state is estimated based on the multi-source asynchronous redundant information; it is further superimposed with the non-sensorized information of the high-precision map sensor, and finally the estimation of the state quantity in the driving environment space is completed.

Further, the at least part of the perception targets may include connected vehicles and other obstacles other than the connected vehicles; wherein, the sensorized information may include connected vehicle detection information for detecting connected vehicles and information on connected vehicles other than the connected vehicles. Non-connected vehicle detection information for obstacles other than connected vehicles.

Further, the non-networked vehicle detection information may include one or more of the following: dynamic obstacle information for detecting dynamic obstacles other than the networked vehicle; dynamic obstacle information for detecting dynamic obstacles on the map; Detected map detection information.

Wherein, the connected vehicle may be a vehicle located in the same vehicle network as the autonomous vehicle; the dynamic obstacle may be a movable obstacle, such as a truck, an animal, a person, etc. that are not connected to the vehicle network. ; The static obstacle can be an immovable obstacle, such as a building, a traffic light, and the like.

Specifically, in the joint perception system, the original map data can be selectively loaded according to perception requirements and vehicle status, and the map data can be described with precision to form an output consistent with common sensor data. This process is called map Sensorization of data.

为地图中关于联网车辆i的数据对该联网车辆状态形成的观测，称为地图-联网车辆观测。

为该观测对应的时间集合。可以定义高精地图传感器联网车辆观测集合：Specifically, the prior information about other connected vehicles in the map is regarded as the observation data of connected vehicles,

It is an observation formed by the data about the connected vehicle i in the map of the state of the connected vehicle, which is called map-connected vehicle observation.

is the time set corresponding to this observation. A collection of HD map sensor networked vehicle observations can be defined:

Wherein, the state of the connected vehicle is used to represent the state information of the connected vehicle that the autonomous vehicle needs to perceive, such as the pose of the connected vehicle (a 6-DOF model, including the origin of the connected vehicle coordinate system in the geographic projection coordinate system) position and angle), size, etc.

为地图中关于联网车辆o的数据对该动态障碍物状态形成的观测，称为地图-动态障碍物观测。

为该观测对应的时间集合。则高精地图传感器动态障碍物观测集合定义为：The prior information about the dynamic obstacles of the non-connected vehicles in the map is regarded as the observation data of the dynamic obstacles.

The observation of the dynamic obstacle state formed by the data about the networked vehicle o in the map is called map-dynamic obstacle observation.

is the time set corresponding to this observation. Then the HD map sensor dynamic obstacle observation set is defined as:

The dynamic obstacle state is used to represent the state information of the dynamic obstacle that the autonomous vehicle needs to perceive, such as the position, size, and category of the dynamic obstacle in the geographic projection coordinate system.

为地图中关于静态障碍物f的数据对该特征状态形成的观测，称为地图-静态障碍物观测。将高精地图传感器静态障碍物观测集合定义为：The prior information about the localization features in the map is regarded as the observation data of static obstacles.

It is the observation formed by the data about the static obstacle f in the map of the feature state, which is called map-static obstacle observation. The HD map sensor static obstacle observation set is defined as:

Wherein, the static obstacle state is used to represent the state information of the static obstacle that the autonomous vehicle needs to perceive, such as the position, size, category, etc. of the static obstacle.

In summary, the high-precision map sensor observations are modeled as a collection of map-static obstacle features, map-connected vehicles, and map-dynamic obstacle observations:

Z ^M = {Z ^MF , Z ^MV , Z ^Mo }

Further, the step of using a map sensor to perform feature extraction on at least a part of the sensing targets in the map to obtain sensorized information may include: determining the position information of the sensing target to be sensed according to the sensing target indication information; According to the accuracy of the map, the covariance of the position information of the at least a part of the perceptual target is calculated and used as the sensorized information.

Specifically, the sensing target to be perceived is determined according to the sensing target indication information, and the sensing target to be perceived may also be referred to as a set of road logical breakpoints. Specifically, the set of road logical breakpoints can be the sensing demand range input by the map sensor, and its corresponding position on the high-precision map (road network position, road segment in the road network, road intersection, and road intersection) is determined according to the positioning information and dynamic and static obstacles. The reference point for dividing the smallest unit of road structure such as the mouth).

Further, the covariance is used to represent sensor noise or perceived uncertainty, and the covariance may be determined by using a conventional appropriate calculation method, which is not limited in this embodiment of the present application.

In other words, the map data can be all the information in the global scope on the map, and the map sensor can generate sensor signals related to the driving task according to the perceived range of the vehicle. Therefore, the original map data needs to be effectively processed. organize.

Starting from the reliability of the map data, according to the perception requirements, the accurate and reliable hierarchical vector high-precision map data is divided into non-sensorized information and sensorized information, and the accurate and reliable real-time map data (road-level and lane-level road network information , other dynamic traffic flow information, traffic event information and decision assistance information) are directly output as non-sensorized information for state estimation in the driving environment space.

Other data is sensorized, output in the form of map sensor signals, and fused with other sensor data in the multi-vehicle joint perception system. This part of the data determines the set of road logical breakpoints related to the driving task through the perceived demand range input by the map sensor. When generating the map sensor signal, firstly, in the positioning feature layer and the dynamic obstacle layer, according to the set of logical breakpoints, all relevant dynamic and static target positions are indexed, and according to the accuracy of the map data, the covariance of the data is determined as the output of the map sensor. Signal.

In the specific implementation of step S13, non-sensorized information other than the above-mentioned sensorized information in the high-precision map data is extracted, superimposed with ordinary sensor perception information and map sensor perception information, and fusion perception is performed under a unified map space-time reference. , and solve the problem through state estimation to obtain the environmental perception results of the self-driving car.

Referring to FIG. 2 , FIG. 2 is a flowchart of a specific implementation manner of step S13 in FIG. 1 . According to the sensorized information, the step of determining the perception result of the autonomous vehicle may include steps S21 to S24, and may further include steps S21 to S26, and each step will be described below.

In step S21, sensor information for detecting the at least a part of the sensing target by one or more sensors is acquired.

Wherein, the sensors may include own vehicle sensors and other vehicle sensors. That is to say, the sensor information may include self-detection information of the self-driving car itself detected by the self-vehicle sensor and external detection information of each other vehicle's sensor detecting other objects (including the current self-driving car) other than the self-driving car. information.

Specifically, the sensor information may include combined navigation observations about its own state, and external sensor observations of other targets, wherein the externally observed objects may be other intelligent networked vehicles, dynamic obstacles, or map data. Included or to-be-updated targets.

第i辆联网车辆组合定位数据时刻的集合为

则感知空间中直接对车辆自身状态进行观测的组合定位观测可以建模为：Further, in the step of detecting the self-driving car itself to obtain self-detection information, the observation data generated by the i-th networked vehicle integrated positioning system on the state of its own vehicle at time k _i can be described as:

The set of combined positioning data moments of the i-th connected vehicle is:

Then the combined positioning observation that directly observes the vehicle's own state in the perception space can be modeled as:

Wherein, the perception space is used to represent the environmental space in which the sensor can detect the target.

Further, the external detection information may include one or more of the following: Internet-connected vehicle detection information for detecting other Internet-connected vehicles in the Internet of Vehicles; dynamic obstacles for detecting dynamic obstacles other than the Internet-connected vehicle. obstacle information; map detection information for detecting static obstacles and road information on the map; newly added static obstacle detection information for detecting static obstacles not recorded on the map; The newly added road detection information for detecting the road information of the map.

Specifically, through the Internet of Vehicles, the data of the external target sensors of other vehicles is obtained to make up for the lack of the perception ability of bicycles. Considering the actual scene, in addition to the observations of dynamic obstacles and static non-map targets, the vehicle-mounted external target sensor data also includes observations of other connected vehicles, positioning features, and road information.

为静态非地图目标的索引，

为第i辆联网车辆在时刻k_i对另一辆网联车

的观测数据，该观测对应的时刻集合为

对应的集合定义为：Further, in the step of detecting other connected vehicles in the Internet of Vehicles to obtain the detection information of the connected vehicles, the observation of other vehicles by the connected vehicle may be called vehicle-vehicle observation, and denoted as vehicle-to-vehicle observation.

is the index of the static non-map object,

For the i-th connected vehicle at time k _i, to another connected vehicle

The observation data of , the time set corresponding to the observation is

The corresponding set is defined as:

其对应的时刻集合为

则车-障碍物观测集合为：In the step of detecting dynamic obstacles other than the connected vehicle to obtain dynamic obstacle information, it can be assumed that a sensor of a connected vehicle i observes a dynamic obstacle o at time k _i , and its data describes for

Its corresponding time set is

Then the vehicle-obstacle observation set is:

为第i辆联网车辆在时刻k_i对定位特征f的观测数据，该观测对应的时刻集合为

对应的集合定义为：In the step of detecting the static obstacles and road information on the map to obtain the map detection information, the observation of the positioning feature by the connected vehicle can be called the vehicle-feature observation, and f is the index of the static non-map target,

is the observation data of the _i -th networked vehicle on the positioning feature f at time ki, and the time set corresponding to the observation is

The corresponding set is defined as:

为第i辆联网车辆在时刻k_i对静态非地图目标s的观测数据，该观测对应的时刻集合为

则对应的观测集合定义为：In the step of detecting static obstacles not recorded in the map to obtain new static obstacle detection information, the observation of the static non-map target may be called vehicle-static non-map target observation, and s is the static non-map target observation. the index of the map object,

is the observation data of the _i -th networked vehicle on the static non-map target s at time ki, and the time set corresponding to the observation is

Then the corresponding observation set is defined as:

为第i辆联网车辆在时刻k_i对道路信息r的观测数据，该观测对应的时刻集合为

对应的集合定义为：In the step of detecting the road information not recorded in the map to obtain the newly added road detection information, the observation of the road information by the connected vehicle is called vehicle-road information observation, and r is the index of the road information,

is the observation data of the road information r of the _i -th connected vehicle at time ki, and the time set corresponding to the observation is

The corresponding set is defined as:

Combining the above information, the external target sensor observations in the perception space can be modeled as:

Z ^V = {Z ^VO , Z ^VS , Z ^VV , Z ^VF , Z ^VR }

In step S22, the sensor information of the at least a part of the sensing targets and the sensorized information are fused to obtain fusion information of each sensing target in the at least a part of the sensing targets.

It should be noted that the technology for fusing the sensor information of the at least a part of the perceptual target and the sensorized information may be performed by using an appropriate conventional technology, which is not limited in this embodiment of the present application.

In step S23, according to the fusion information, an observation quantity set of each sensing target in the at least a part of the sensing targets is determined.

可以包括联网车辆

和其他动态障碍物

对于动态目标构建状态集合X^H，动态目标状态集合可以表示为：Wherein, according to the sensing method in the embodiment of the present application, the sensing information of the dynamic and static targets and the road traffic information can be determined according to the sensing requirement of the intelligent vehicle. where dynamic target

Can include connected vehicles

and other dynamic obstacles

For the dynamic target construction state set X ^H , the dynamic target state set can be expressed as:

X ^H ={X ^V ,X ^O }

包含地图中已经包括的定位特征

以及静态非地图目标

也即地图待更新目标。对于静态目标构建状态集合X^Z，静态目标状态集合可以表示为：Static objects such as traffic signs, lane lines, and cones in traffic scenes provide reference information for autonomous vehicle positioning. In addition, static targets within the road edge need to be considered in the decision system. Static targets in the environment

Include location features already included in the map

and static non-map targets

That is, the target to be updated on the map. For a static target build state set X ^Z , the static target state set can be expressed as:

X ^Z ={X ^F ,X ^S }

包括道路路网信息

车道路网信息

动态交通信息

以及决策辅助信息

对于道路信息构建状态集合X^R，道路信息状态集合可以表示为：road information state set

Include road network information

road network information

Dynamic Traffic Information

and decision aids

For the road information construction state set X ^R , the road information state set can be expressed as:

X ^R = {X ^WL , X ^w , X ^B , X ^P }

To sum up, the target in the driving environment space can be composed of dynamic target, static target and road information, and its mathematical definition is:

The corresponding perception result of the driving car, that is, the observation set can be expressed as:

X ^D ={X ^H ,X ^Z ,X ^R }

In the specific implementation of step S24, a maximum likelihood estimation algorithm may be used to determine the state quantity estimation of each sensing target in the at least a part of sensing targets according to the observation quantity set of each sensing target in the at least a part of sensing targets.

Specifically, in the various perceptual inputs of the joint perception system, there are generally observations of the same target from different sources, and there are contradictions between them. To obtain consistent perceptual results, maximum likelihood estimation is performed. Because the joint perception problem needs to describe the state under a unified map space-time reference, it can be attributed to maximizing the likelihood probability of all the observed quantities of the system about the state quantity. According to the maximum likelihood estimation (Maximum Likelihood Estimation, MLE) theory, the driving environment space state estimator maximizes the probability of the observations conditioned on the set of observations.

In this embodiment of the present invention, by acquiring sensor information of one or more sensors for detecting the at least a part of the perceptual targets, the fusion information and the observation quantity set of each perceptual target in the at least a part of the perceptual targets are obtained, and then the set of the perceptual objects is determined. The estimation of the state quantity of each sensing target in the at least a part of the sensing targets can realize the determination of the sensing result on the basis of the information fusion of the map sensor and the sensor, and further improve the sensing accuracy.

Further, the following formula can be used to determine the state quantity estimation of each sensing target in the at least a part of the sensing targets according to the observation quantity set of each sensing target in the at least a part of the sensing targets:

用于表示极大似然意义下驾驶环境空间状态参数的估计结果，X^D用于表示驾驶环境空间中一定时间深度内的观测量集合，Z^p用于表示智能汽车可获取的异步异构的感知空间中的测量集合，P(Z^p|X^D)是在给定状态参数X^D下观测量集合Z^p的条件概率。in,

It is used to represent the estimation results of the state parameters of the driving environment space in the sense of maximum likelihood, X ^D is used to represent the set of observations within a certain time depth in the driving environment space, and Z ^p is used to represent the asynchronous and heterogeneous data available to smart cars. The set of measurements in perceptual space, P(Z ^p |X ^D ) is the conditional probability of observing the set of measurements Z ^p given the state parameter X ^D .

Assuming that the likelihood probabilities of different sensors and different moments are independent of each other, the conditional probability can be further written as the joint product of the likelihood probabilities of all observations in a period of time.

Assuming that the noise follows a Gaussian distribution, the maximizing likelihood problem can be transformed into the following least squares optimization problem. In this formula, the residual is the difference between the actual observation and the prediction of the perception result by the system state. In the state estimator of the perception container, the weighted sum of various observation residuals is minimized by optimizing the value of the state quantity.

Observed residuals:

The above formula gives the mathematical method for estimating the state X ^D in the driving environment space by the state estimator in the map-aware container. It should be noted that this formula adjusts the weight of sensor data of different precisions in the optimization problem through the inverse matrix of the covariance matrix of each sensor noise, so as to organically fuse the vehicle sensors and map sensors in the entire joint perception system. data, and finally form a consistent perception result.

In the embodiment of the present invention, the maximum likelihood estimation algorithm is used to estimate the perception result of the self-driving car, which can effectively improve the accuracy of the estimation result.

In the specific implementation of step S25, the state quantity estimation of each sensing target in the at least a part of the sensing targets is superimposed with the non-sensorized information to obtain the sensing result of the autonomous vehicle.

Further, the non-sensorized information may include one or more of the following: road information, newly added traffic flow information, traffic event information, and decision assistance information.

Specifically, the non-sensorized information may include road-level and lane-level road network information, dynamic traffic flow information, traffic event information, decision assistance information, and the like.

It should be noted that, in the step of superimposing the state quantity estimates of each of the at least a part of the perceptual objects with the non-sensorized information, non-sensor information (road network, traffic information, etc.) and sensors (including The information of the actual sensor and the map virtual sensor) are unifiedly expressed in the global coordinate system of the high-precision map, which realizes the consistency of the spatial expression and can be directly superimposed.

In this embodiment of the present invention, non-sensorized information other than the sensorized information may also be extracted from the map, and further, the state quantity estimation of each of the at least a part of the perceptual objects is associated with the non-sensorized information. The information is superimposed to obtain the perception result of the self-driving car, so that other information of the high-precision map can be further added on the basis of the detection of various sensors, and the overall effect of the perception can be further improved.

In the embodiment of the present invention, by acquiring sensing target indication information, the sensing target indication information is used to indicate the environmental state information that the self-driving car needs to perceive; acquiring a high-precision map, establishing a map perception container in the map Feature extraction is performed on at least a portion of the perceptual objects to obtain sensorized information, thereby generating a signal in the same format as a common sensor based on the map raw data. In addition, considering that some information in the high-precision map data has far more confidence than the on-board sensors, it is directly superimposed on the perception results. Finally, the multi-vehicle joint perception information (including ordinary sensor perception information, map sensor perception information, and other non-sensorized data in high-precision maps) is fused under the unified map space-time reference to determine the perception result of the autonomous vehicle; wherein , the positioning error of the map is within a preset error range. The present invention can effectively eliminate the perception blind area and improve the perception precision.

Referring to FIG. 3 , FIG. 3 is a schematic structural diagram of an apparatus for designing a high-precision map sensing container for multi-vehicle joint sensing according to an embodiment of the present invention. The high-precision map perception container design device for multi-vehicle joint perception may include:

an information acquisition module 31, configured to acquire sensing target indication information, where the sensing target indication information is used to indicate the environmental state information that the autonomous vehicle needs to perceive;

The sensorized information determination module 32 is used to obtain a high-precision map, establish a map perception container, and perform feature extraction on at least a part of the perception targets in the map to obtain sensory information;

The perception result solving module 33 is used to extract the non-sensorized information in the high-precision map data except the above-mentioned sensorized information, superimpose with the common sensor perception information and map sensor perception information, and perform fusion perception under the unified map space-time reference , and solve the problem through state estimation to obtain the environmental perception results of the self-driving car.

The high-precision map is a map with a positioning error within a preset error range.

For the principle, specific implementation and beneficial effects of the multi-vehicle joint perception-oriented high-precision map perception container design device, please refer to the relevant description of the multi-vehicle joint perception-oriented high-precision map perception container design method described above. Repeat.

Referring to FIG. 4 , FIG. 4 is a schematic structural diagram of a sensing system based on a multi-vehicle joint sensing-oriented high-precision map sensing container design method according to an embodiment of the present invention.

As shown in Figure 4, the solid line box is used to represent the entity module, and the dotted line box is used to represent the determined information.

First, sensor information determined by the plurality of sensors S ₁ , S ₂ , S ₃ to S ₄ is collected.

The map sensor M ^S is used to determine the map information M, and the sensing target indication information Pr is determined to output ^sensorized information Z ^M .

Further, the driving environment space builder DC is used to fuse the sensor information and the ^sensorized information of the at least a part of the perceptual targets, so as to obtain the fusion information of each perceptual target in the at least a part of the perceptual targets, and according to the fusion information , determining an observation set of each sensing target in the at least a part of the sensing targets.

In a specific implementation, the ^DC can also allocate the state quantities in the driving environment space according to the expression results of the multi-source sensors in the map coordinate system.

More specifically, Dc associates data according to the expression results of multi-source sensors in the geographic projection coordinate system, determines the set of observations that need to be estimated, allocates the state quantities in the driving environment space, and completes the construction of the driving environment space. .

And then adopt the driving environment space state estimator ^ME based on the maximum likelihood estimation algorithm, according to the observation quantity set of each sensing target in the at least a part of the sensing target, determine the state quantity estimation of each sensing target in the at least a part of the sensing target

Specifically, ^ME is a driving environment space state estimator, which realizes the estimation of state quantities in the driving environment space based on multi-source heterogeneous and asynchronous redundant data.

Further, a map sensor MS is used to extract non- ^sensorized information X ₀ in the map except the sensorized information.

与所述非传感器化信息X₀叠加，以得到所述自动驾驶汽车的感知结果

estimating the state quantity of each sensing target in the at least a part of sensing targets

Superimposed with the non-sensorized information X ₀ to obtain the perception result of the self-driving car

For more details of the sensing system shown in FIG. 4 , the execution can be performed with reference to the foregoing description, which will not be repeated here.

In an embodiment of the present invention, a storage medium is also provided, on which a computer program is stored, and the computer program executes the steps of the above method when the computer program is run by a processor. The storage medium may be a computer-readable storage medium, for example, may include non-volatile memory (non-volatile) or non-transitory (non-transitory) memory, and may also include optical disks, mechanical hard disks, solid-state disks, and the like.

Specifically, in this embodiment of the present invention, the processor may be a central processing unit (central processing unit, CPU for short), and the processor may also be other general-purpose processors, digital signal processors (digital signal processor, DSP for short) , an application specific integrated circuit (ASIC for short), an off-the-shelf programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and the like. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

It should also be understood that the memory in the embodiments of the present application may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory may be a read-only memory (ROM for short), a programmable read-only memory (PROM for short), an erasable PROM for short (EPROM) , Electrically Erasable Programmable Read-Only Memory (electrically EPROM, EEPROM for short) or flash memory. The volatile memory may be random access memory (RAM for short), which is used as an external cache memory. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic Random access memory (synchronous DRAM, referred to as SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, referred to as DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, referred to as ESDRAM), synchronous connection Dynamic random access memory (synchlink DRAM, referred to as SLDRAM) and direct memory bus random access memory (direct rambus RAM, referred to as DR RAM).

In an embodiment of the present invention, a terminal is also provided, including a memory and a processor, the memory stores a computer program that can run on the processor, and the processor executes the above-mentioned computer program when the processor runs the computer program. steps of the method. The terminals include but are not limited to automobiles, automobile central control equipment, terminal equipment externally connected to or integrated in the automobile, mobile phones, computers, tablet computers and other terminal equipment.

Although the present invention is disclosed above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be based on the scope defined by the claims.

Claims (10)

1. a high-precision map perception container design method for multi-vehicle joint perception, is characterized in that, comprises the following steps: acquiring sensing target indication information, where the sensing target indication information is used to indicate the environmental state information that the autonomous vehicle needs to perceive; Obtain a high-precision map, establish a map-aware container, and perform feature extraction on at least a part of the perception targets in the map to obtain sensorized information; Extract the non-sensorized information in the high-precision map data except the above-mentioned sensorized information, superimpose it with ordinary sensor perception information and map sensor perception information, and perform fusion perception under the unified map space-time reference, and solve the problem through state estimation. Environmental perception results of driving a car; The high-precision map is a map with a positioning error within a preset error range. 2 . The high-precision map perception container design method for multi-vehicle joint perception according to claim 1 , wherein the at least part of the perception targets include connected vehicles and other obstacles other than the connected vehicles; 2 . Wherein, the sensorized information includes connected vehicle detection information for detecting connected vehicles and non-connected vehicle detection information for detecting obstacles other than the connected vehicle. 3. The high-precision map perception container design method for multi-vehicle joint perception according to claim 2, wherein the non-networked vehicle detection information comprises one or more of the following: dynamic obstacle information for detecting dynamic obstacles other than the connected vehicle; Map detection information for detecting static obstacles on the map. 4 . The high-precision map sensing container design method for multi-vehicle joint sensing according to claim 1 , wherein, using a map sensor to perform feature extraction on at least a part of sensing targets in the map to obtain sensorized information comprising: 5 . : Determine the location information of the sensing target to be sensed according to the sensing target indication information; According to the accuracy of the map, the covariance of the position information of the at least a part of the perceptual target is calculated and used as the sensorized information. 5 . The high-precision map perception container design method for multi-vehicle joint perception according to claim 1 , wherein the extraction of non-sensorized information other than the above-mentioned sensorized information in the high-precision map data is the same as that of ordinary The sensor perception information and the map sensor perception information are superimposed, and the fusion perception is carried out under the unified map space-time reference, and the environmental perception results of the autonomous vehicle are obtained by solving the problem through state estimation: acquiring sensor information for detecting the at least a part of the sensing target by one or more sensors; Fusion of the sensor information of the at least a part of the perceptual targets and the sensorized information to obtain the fusion information of each perceptual target in the at least a part of the perceptual targets; determining, according to the fusion information, an observation set of each perceptual target in the at least a part of the perceptual targets; A maximum likelihood estimation algorithm is used to determine the state quantity estimation of each sensing target in the at least a part of sensing targets according to the observation quantity set of each sensing target in the at least a part of sensing targets. 6. The high-precision map perception container design method for multi-vehicle joint perception according to claim 5, wherein the extraction of non-sensorized information other than the sensorized information in the high-precision map data is the same as that of ordinary The sensor perception information and the map sensor perception information are superimposed, and the fusion perception is carried out under the unified map space-time reference, and the environmental perception results of the autonomous vehicle are obtained by solving the problem through state estimation. The results also include: The state quantity estimation of each sensing target in the at least a part of the sensing targets is superimposed with the non-sensorized information to obtain a sensing result of the autonomous vehicle. 7 . The high-precision map perception container design method for multi-vehicle joint perception according to claim 5 , wherein the following formula is used to determine the set of observations of each perception target in the at least a part of the perception targets. 8 . Describe the state quantity estimation of each perceptual target in at least a part of perceptual targets:

in,

It is used to represent the estimation results of the state parameters of the driving environment space in the sense of maximum likelihood, X ^D is used to represent the set of observations within a certain time depth in the driving environment space, and Z ^p is used to represent the asynchronous and heterogeneous data available to smart cars. The set of measurements in perceptual space, P(Z ^p |X ^D ) is the conditional probability of observing the set of measurements Z ^p given the state parameter X ^D . 8. A high-precision map perception container design device for multi-vehicle joint perception, characterized in that it comprises: an information acquisition module, configured to acquire sensing target indication information, where the sensing target indication information is used to indicate environmental state information that the autonomous vehicle needs to perceive; The sensorized information determination module is used to obtain a high-precision map, establish a map perception container, and perform feature extraction on at least a part of the perception targets in the map to obtain sensory information; The perception result solving module is used to extract the non-sensorized information other than the above-mentioned sensorized information in the high-precision map data, superimpose it with the ordinary sensor perception information and map sensor perception information, and perform fusion perception under the unified map space-time reference. Solve the problem through state estimation to obtain the environmental perception results of the autonomous vehicle; The high-precision map is a map with a positioning error within a preset error range. 9. A storage medium on which a computer program is stored, wherein the computer program executes the design of a high-precision map perception container for joint perception of multiple vehicles according to any one of claims 1 to 7 when the computer program is run by a processor steps of the method. 10. A terminal comprising a memory and a processor, wherein a computer program that can be run on the processor is stored on the memory, wherein the processor executes claims 1 to 7 when running the computer program The steps of any one of the high-precision map perception container design methods for multi-vehicle joint perception.

Images