Performance evaluation method, system and product for automatic driving

Abstract

The embodiment of the invention provides an automatic driving performance evaluation method, system and product, and the method comprises the steps: determining the label data of various label types in driving scene data according to various preset label types; according to label data of various label types in the driving scene data and the basic tree-shaped mathematical model, a driving scene tree-shaped mathematical model is constructed, and each leaf node in the basic tree-shaped mathematical model is composed of various preset label types; determining the complexity of the driving scene according to the driving scene tree-shaped mathematical model; determining the value of an evaluation index according to the driving scene data; and determining an automatic driving performance score according to the value of each evaluation index and the complexity of the driving scene. The objective of the invention is to improve performance evaluation accuracy of automatic driving.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and specifically to a performance evaluation method, system and product for autonomous driving. Background Art

[0002] The ultimate goal of high-level autonomous driving is for autonomous driving algorithms to gradually evolve towards anthropomorphism during the iteration process. However, the current evaluation methods mainly evaluate some basic driving capabilities, and the performance of autonomous driving algorithms is not accurate enough. Summary of the invention

[0003] In view of this, the embodiments of the present application provide a method, system and product for evaluating the performance of autonomous driving, aiming to improve the accuracy of the performance evaluation of autonomous driving.

[0004] A first aspect of an embodiment of the present application provides a performance evaluation method for autonomous driving, the method comprising:

[0005] Determine label data of various label types in the driving scene data according to various preset label types;

[0006] According to the label data of various label types in the driving scene data and the basic tree-like mathematical model, a driving scene tree-like mathematical model is constructed, wherein each leaf node in the basic tree-like mathematical model is composed of various preset label types;

[0007] Determine the complexity of the driving scenario based on the tree-like mathematical model of the driving scenario;

[0008] Determine the value of the evaluation index based on the driving scenario data;

[0009] The autonomous driving performance score is determined based on the values ​​of various evaluation indicators and the complexity of the driving scenario.

[0010] Optionally, a driving scene tree-like mathematical model is constructed according to the label data of various label types in the driving scene data and the basic tree-like mathematical model, including:

[0011] According to various label types in the driving scene data, determining the node positions of various label types in the basic tree mathematical model;

[0012] The label data of various label types in the driving scene data are placed into their corresponding node positions to obtain a tree-like mathematical model of the driving scene.

[0013] Optionally, a basic tree-like mathematical model is constructed, including:

[0014] Determine the natural driving scene node as the root node of the tree mathematical model;

[0015] Determine each preset label class as a child node of the tree-like mathematical model;

[0016] According to the affiliation between the tag classes, the parent-child node relationship between the tag classes is determined;

[0017] Identify various tag types as leaf nodes under the tag class;

[0018] According to the determined root nodes, child nodes, parent-child node relationships and leaf nodes, a basic tree-like mathematical model is constructed.

[0019] Optionally, a basic tree-like mathematical model is constructed according to the determined root node, child nodes, parent-child node relationships, and leaf nodes, including:

[0020] Constructing a first tree-like mathematical model according to the determined root node, child nodes, parent-child node relationships, and leaf nodes;

[0021] According to the driving scenario type, the evaluation weights of the first child nodes directly belonging to the root node in the first tree-like mathematical model are preset to obtain a basic tree-like mathematical model.

[0022] Optionally, determining the complexity of the driving scene according to the tree-like mathematical model of the driving scene includes:

[0023] Determine the complexity values ​​corresponding to the label data of each node in the driving scene tree-like mathematical model according to the driving scene tree-like mathematical model and the target mapping relationship, wherein the target mapping relationship records the mapping relationship between different label data of each node in the driving scene tree-like mathematical model and different complexity values;

[0024] The complexity of the driving scenario is determined according to the complexity values ​​corresponding to the label data of all nodes in the determined tree-like mathematical model of the driving scenario.

[0025] Optionally, determining the complexity of the driving scene according to the tree-like mathematical model of the driving scene includes:

[0026] According to the driving scene tree mathematical model and the target mapping relationship, determining the complexity value corresponding to the label data of each node in the driving scene tree mathematical model;

[0027] Determine a target complexity value of the node according to the complexity value corresponding to the label data of the node and the evaluation weight of the first child node to which the node belongs;

[0028] The complexity of the driving scenario is determined according to the target complexity values ​​of all nodes in the determined driving scenario tree mathematical model.

[0029] Optionally, the autonomous driving performance score is determined based on the values ​​of various evaluation indicators and the complexity of the driving scenario, including:

[0030] The automatic driving performance score is obtained by bringing the determined values ​​of various evaluation indicators and the complexity of the driving scene into the driving performance evaluation algorithm for calculation. The driving performance evaluation algorithm expression is:

[0031] S=C(ω1I1+ω2I2+...+ω n I n )

[0032] Among them, S is the autonomous driving performance score, C is the complexity of the driving scene, ω n is the weight of the nth evaluation index, I n is the value of the nth evaluation index.

[0033] Optionally, when the evaluation index is a safety evaluation index, determining the value of the evaluation index according to the driving scenario data includes:

[0034] The relative distance between the self-vehicle and the target vehicle in the driving scene data and their respective speeds are brought into the safety assessment algorithm for calculation to obtain the value of the safety evaluation index. The safety assessment algorithm expression is:

[0035]

[0036] Among them, U is the safety evaluation index, U x is the vertical risk, U y is the horizontal risk, k lon is the virtual longitudinal spring stiffness, x is the relative distance between the ego vehicle and the target vehicle, x0 is the expected relative distance between the ego vehicle and the target vehicle, c lpn is the longitudinal virtual damping coefficient, v p is the target vehicle speed, v x is the vehicle speed, c lat is the lateral virtual damping coefficient, v y is the lateral speed of the vehicle, y is the lateral offset of the vehicle from the center line of the current lane, k lat is the virtual lateral spring stiffness, U x,max To customize the maximum vertical value and U y,max It is the custom horizontal maximum value.

[0037] Optionally, when the evaluation index is an efficiency evaluation index, determining the value of the evaluation index according to the driving scenario data includes:

[0038] The value of the efficiency evaluation index is obtained by bringing the lane maximum speed limit and the vehicle speed in the driving scenario data into the efficiency evaluation algorithm for calculation. The efficiency evaluation algorithm expression is:

[0039]

[0040] Among them, I e is the efficiency evaluation index, v x is the vehicle speed, v ego,x The maximum speed limit for the lane.

[0041] Optionally, when the evaluation index is a comfort evaluation index, determining the value of the evaluation index according to the driving scenario data includes:

[0042] The value of the comfort evaluation index is obtained by bringing the vehicle acceleration in the driving scene data into the comfort evaluation algorithm for calculation. The expression of the comfort evaluation algorithm is:

[0043]

[0044]

[0045] Among them, I c is the comfort evaluation index, a index is the vehicle acceleration index, a lon is the longitudinal acceleration, a lat is the lateral acceleration, t is the sampling frequency, i is the i-th sampling, N is the number of data points sampled per unit time, jerk index is the acceleration index, jerk lon is the longitudinal acceleration, jerk lat is the lateral acceleration, is the value of the vehicle acceleration, is the human body’s sensitive acceleration value, is the value of the vehicle's acceleration, It is the sensitive acceleration value of human body.

[0046] Optionally, when the evaluation index is an anthropomorphic evaluation index, determining the value of the evaluation index according to the driving scene data includes:

[0047] The value of the anthropomorphic evaluation index is obtained by bringing the vehicle trajectory in the driving scene data into the anthropomorphic evaluation algorithm for calculation. The expression of the anthropomorphic evaluation algorithm is:

[0048]

[0049] D(i,j)=(x i +y i ) 2 +min([D(i-1,j),D(i,j-1),D(i-1,j-1)]);

[0050] Among them, the actual trajectory sequence X={x1,x2,x3,…,x n}, x i is the actual coordinate position of the vehicle at the i-th moment, and the trajectory sequence to be evaluated Y = {y1, y2, y3, …, y m},y i is the coordinate position of the vehicle to be evaluated at the i-th moment, and n and m represent the lengths of the two sequences respectively.

[0051] A second aspect of an embodiment of the present application provides a performance evaluation system for autonomous driving, the system comprising:

[0052] A label data determination module, used to determine label data of various label types in the driving scene data according to various preset label types;

[0053] A driving scene tree-like mathematical model construction module is used to construct a driving scene tree-like mathematical model according to label data of various label types in the driving scene data and a basic tree-like mathematical model, wherein each leaf node in the basic tree-like mathematical model is composed of various preset label types;

[0054] A driving scenario complexity determination module, used to determine the complexity of the driving scenario according to a tree-like mathematical model of the driving scenario;

[0055] An evaluation index value determination module is used to determine the value of the evaluation index according to the driving scene data;

[0056] The autonomous driving performance score determination module is used to determine the autonomous driving performance score based on the values ​​of various evaluation indicators and the complexity of the driving scenario.

[0057] A third aspect of an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and running on the processor. When the computer program is executed by the processor, it implements the steps in a performance evaluation method for autonomous driving as described in the first aspect of the present application.

[0058] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, characterized in that a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the performance evaluation method of autonomous driving as described in the first aspect of the present application are implemented.

[0059] The performance evaluation method of autonomous driving provided in this application has the following advantages:

[0060] The embodiment of the present application provides a method for evaluating the performance of an autonomous driving system. First, according to various preset label types, the label data of various label types in the driving scene data are determined; according to the label data of various label types in the driving scene data and the basic tree-like mathematical model, a tree-like mathematical model of the driving scene is constructed, and each leaf node in the basic tree-like mathematical model is composed of various preset label types; according to the tree-like mathematical model of the driving scene, the complexity of the driving scene is determined; according to the driving scene data, the value of the evaluation index is determined; according to the value of each evaluation index and the complexity of the driving scene, the autonomous driving performance score is determined. Therefore, the driving scene of the present application will also affect the performance of the autonomous driving algorithm. When evaluating the performance of the autonomous driving algorithm, not only the performance of the algorithm under multiple evaluation indicators is considered, but also differentiated driving scenarios are considered in the evaluation process to evaluate the performance of the autonomous driving algorithm, thereby effectively improving the accuracy of the performance evaluation of the autonomous driving algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0062] Figure 1 A flowchart of a method for evaluating the performance of an autonomous driving system according to an embodiment of the present application;

[0063] Figure 2 A schematic diagram of a basic tree-like mathematical model in a performance evaluation method for autonomous driving shown in one embodiment of the present application;

[0064] Figure 3 Another flow chart of a method for evaluating the performance of an autonomous driving system according to an embodiment of the present application;

[0065] Figure 4 A schematic diagram of an autonomous driving performance evaluation system is shown as an embodiment of the present application. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0067] refer to Figure 1 , Figure 1 The following is a flow chart of a method for evaluating the performance of an autonomous driving system according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0068] Step S1: Determine tag data of various tag types in driving scene data according to various preset tag types.

[0069] In this embodiment, the present application is concerned with the perception of the driving environment in the process of controlling the vehicle to perform autonomous driving, such as collecting driving scene data through image sensors, laser radars and other devices. In the present application, for the driving scene data processed by the autonomous driving algorithm, multiple label types are first pre-set, and then when evaluating the performance of the autonomous driving algorithm in the later stage, the driving scene data perceived by the autonomous driving algorithm in the process of controlling the vehicle to perform autonomous driving is labeled according to various preset label types, and the values ​​of the driving scene data on the various label types are determined. Among them, the multiple label types include but are not limited to time, location, weather and lighting, road type, road surface conditions, road characteristics, lane type, traffic congestion level, signal light status, and the type, motion state, behavior intention and initial direction of traffic participants. For example, for the time label type, the perception time of the driving scene data is determined by analyzing and processing the driving scene data. After the perception time is determined, the perception time is the label data of the time label type; for the road type label type, the driving scene data is analyzed and processed to determine whether the road in the driving scene data is a highway, a city expressway, an urban road, etc. After it is determined that the road type is a highway, the highway is the label data of the road type label type.

[0070] In this embodiment, when the label type is time, the corresponding specific label data includes at least daytime, early morning, evening, and night; when the label type is location, the corresponding specific label data includes at least longitude, latitude, and elevation; when the label type is weather, the corresponding specific label data includes at least cloudy, light snow, rain, and fog; when the label type is weather, the corresponding specific label data includes at least bright and dim; when the label type is traffic light status, the corresponding specific label data includes at least red, yellow, and green; when the label type is traffic congestion level, the corresponding specific label data includes at least strong, medium, and weak; when the label type is road type, the corresponding specific label data includes at least expressways, urban expressways, urban roads, and urban intersections / intersections; when the label type is road surface conditions, the corresponding specific label data includes at least dry, wet, waterlogged, muddy, and potholes; when the label type is road features, the corresponding specific label data includes at least single lane, double lane, three lanes, and multiple lanes; When the label type is the lane type, the corresponding specific label data includes at least motor vehicle lane and non-motor vehicle lane; when the label type is the vehicle object type in the traffic participant object class, the corresponding specific label data includes at least large truck, small car and pickup truck; when the label type is the person object type in the traffic participant object class, the corresponding specific label data includes at least cyclist, adult and child; when the label type is other object types in the traffic participant object class, the corresponding specific label data includes at least animal and wheelchair; the label types under the motion state subclass in the traffic participant object class include at least speed, acceleration, lateral relative distance and longitudinal relative distance; when the label type is behavior intention, the corresponding specific label data includes at least cut-in and cut-out; when the label type is initial direction, the corresponding specific label data includes at least forward and left; when the label type is lane line, the corresponding specific label data includes at least dotted line, solid line and yellow; when the label type is lane arrow, the corresponding specific label data includes at least straight arrow, left turn arrow and right turn arrow.

[0071] Step S2: constructing a driving scene tree-like mathematical model according to the tag data of various tag types in the driving scene data and the basic tree-like mathematical model, wherein each leaf node in the basic tree-like mathematical model is composed of various preset tag types.

[0072] In this embodiment, the present application pre-constructs a basic tree-like mathematical model, in which each leaf node is composed of various preset label types, one label type corresponds to one leaf node, and after obtaining the specific label data of various label types in the driving scene data through step S1, each label data is filled into the corresponding leaf node in the basic tree-like mathematical model to obtain the driving scene tree-like mathematical model. For example, the preset label type involves the road type, so there is a leaf node belonging to the label type of the road type in the basic tree-like mathematical model. When it is determined through step S1 that the label data of the road type is a highway, the highway is filled into the leaf node belonging to the label type of the road type in the basic tree-like mathematical model. After all the label data obtained by analyzing and processing the driving scene data through step S1 are filled into the corresponding leaf nodes, the driving scene tree-like mathematical model is obtained.

[0073] Step S3: Determine the complexity of the driving scenario according to the tree-like mathematical model of the driving scenario.

[0074] In this embodiment, after the tree-like mathematical model of the driving scene is obtained through step S2, the complexity of each leaf node is determined based on the specific label data recorded in each leaf node in the tree-like mathematical model of the driving scene. An optional implementation is: for each leaf node (that is, each label type) with different label data, a corresponding complexity is set for each label data by expert scoring, and then a mapping relationship between different label data and complexity of different leaf nodes is formed. After the tree-like mathematical model of the driving scene is obtained, based on the specific label data recorded in each leaf node in the tree-like mathematical model of the driving scene, the complexity of each leaf node in the tree-like mathematical model of the driving scene is determined from the mapping relationship, and finally based on the complexity of all leaf nodes in the tree-like mathematical model of the driving scene, a driving scene complexity is determined. An optional implementation is to sum the complexity of all leaf nodes in the tree-like mathematical model of the driving scene to obtain a driving scene complexity. For example, for weather label types, different label data include cloudy, dense fog, rain and snow, hail, typhoon, etc. According to expert scoring, the complexity value is set to 0.2 for cloudy, 0.6 for dense fog, 0.4 for rain and snow, 0.6 for hail, 0.9 for typhoon, etc.

[0075] Step S4: Determine the value of the evaluation index according to the driving scenario data.

[0076] In this embodiment, the various evaluation indicators at least include: safety evaluation indicators, efficiency evaluation indicators, comfort evaluation indicators and anthropomorphic evaluation indicators. The driving scene data perceived during the process of the automatic driving algorithm controlling the vehicle to perform automatic driving is analyzed to determine the performance of the automatic driving algorithm under various evaluation indicators. The better the performance under the corresponding evaluation indicator, the higher the value of the corresponding evaluation indicator, and the worse the performance under the corresponding evaluation indicator, the lower the value of the corresponding evaluation indicator. For example, the driving scene data perceived during the process of the automatic driving algorithm controlling the vehicle to perform automatic driving is analyzed to determine that the better the performance of the automatic driving algorithm under the safety evaluation indicator, the higher the value of the safety evaluation indicator; the driving scene data perceived during the process of the automatic driving algorithm controlling the vehicle to perform automatic driving is analyzed to determine that the worse the performance of the automatic driving algorithm under the comfort evaluation indicator, the lower the value of the comfort evaluation indicator.

[0077] Step S5: Determine the autonomous driving performance score based on the values ​​of various evaluation indicators and the complexity of the driving scenario.

[0078] In this embodiment, after analyzing the driving scene data perceived by the automatic driving algorithm in the process of controlling the automatic driving of the vehicle through steps S1 to S4 to obtain the values ​​of various evaluation indicators and the complexity of the driving scene, the automatic driving performance score of the automatic driving algorithm is determined based on the values ​​of various evaluation indicators and the complexity of the driving scene. An optional implementation method is to first accumulate the values ​​of various evaluation indicators, and then multiply the accumulated values ​​by the driving scene complexity value to obtain the corresponding automatic driving performance score.

[0079] The embodiment of the present application provides a method for evaluating the performance of an autonomous driving system. First, according to various preset label types, the label data of various label types in the driving scene data are determined; according to the label data of various label types in the driving scene data and the basic tree-like mathematical model, a tree-like mathematical model of the driving scene is constructed, and each leaf node in the basic tree-like mathematical model is composed of various preset label types; according to the tree-like mathematical model of the driving scene, the complexity of the driving scene is determined; according to the driving scene data, the value of the evaluation index is determined; according to the value of each evaluation index and the complexity of the driving scene, the autonomous driving performance score is determined. Therefore, the driving scene of the present application will also affect the performance of the autonomous driving algorithm. When evaluating the performance of the autonomous driving algorithm, not only the performance of the algorithm under multiple evaluation indicators is considered, but also differentiated driving scenarios are considered in the evaluation process to evaluate the performance of the autonomous driving algorithm, thereby effectively improving the accuracy of the performance evaluation of the autonomous driving algorithm.

[0080] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, step S2 may include: determining the node positions of various label types in the basic tree-like mathematical model according to various label types in the driving scene data; placing the label data of various label types in the driving scene data into their respective corresponding node positions to obtain a tree-like mathematical model of the driving scene.

[0081] In this embodiment, a leaf node in the basic tree-like mathematical model corresponds to a label type. After the driving scene data is analyzed and processed in step S1 to obtain all the label data, the label type corresponding to the label data is determined for each label data, and then the leaf node position of the label type in the basic tree-like mathematical model is determined, and finally the label data is filled into the corresponding leaf node position. For example, for the label data highway, the label type corresponding to the label data is determined to be the road type, and then the leaf node position of the road type in the basic tree-like mathematical model is determined to be position A, so the label data of the highway is filled into position A in the basic tree-like mathematical model. After all the label data are filled into the basic tree-like mathematical model, the corresponding driving scene tree-like mathematical model is obtained.

[0082] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, the method further includes constructing a basic tree-like mathematical model, which may include steps S01 to S05:

[0083] Step S01: determine the natural driving scene node as the root node of the tree mathematical model.

[0084] In this embodiment, if Figure 2 As shown, the present application determines the natural driving scene node as the root node of the tree-like mathematical model, and the root node does not record data information and is just an empty node.

[0085] Step S02: Determine each preset label class as a child node of the tree mathematical model.

[0086] In this embodiment, the present application presets multiple label classes, and then divides each label type into a corresponding label class, and determines each label class as a sub-node of the tree-like mathematical model, wherein the sub-node of the tree-like mathematical model refers to the other nodes of the tree-like mathematical model except the root node and the leaf node, and all of them belong to the sub-node of the tree-like mathematical model.

[0087] Step S03: Determine the parent-child node relationship between tag classes according to the affiliation relationship between tag classes.

[0088] In this embodiment, the preset multiple tag classes include tag classes with affiliation, for example, tag class A includes two smaller tag classes a1 and a2. Then, according to the affiliation between the tag classes, the parent-child node relationship between the tag classes is determined, that is, for two tag classes with affiliation, the other tag class belonging to one tag class is determined as the child node of the one tag class.

[0089] Step S04: Determine various tag types as leaf nodes under the tag class.

[0090] In this embodiment, based on the tag classes to which the various tag types are divided, each tag type is determined as a leaf node under its corresponding tag class.

[0091] Step S05: construct a basic tree-like mathematical model according to the determined root nodes, child nodes, parent-child node relationships and leaf nodes.

[0092] In this embodiment, after the root node in the tree-like mathematical model, the child nodes in the tree-like mathematical model, the parent-child node relationship between each child node in the tree-like mathematical model, and the leaf nodes in the tree-like mathematical model are determined through steps S01 to S04, a basic tree-like mathematical model is constructed based on the determined root node in the tree-like mathematical model, the child nodes in the tree-like mathematical model, the parent-child node relationship between each child node in the tree-like mathematical model, and the leaf nodes in the tree-like mathematical model.

[0093] In this embodiment, if Figure 2 As shown, Figure 2 An exemplary basic tree-like mathematical model is shown, in which the autonomous driving scene is determined as the root node; the preset label classes include general classes (where T represents time, L represents location, W represents weather, and S represents light), traffic classes, road classes, traffic participant classes (such as vehicles, pedestrians, etc.), and motion states. Objects 1 and n in the figure are represented for ease of understanding and do not belong to a node in the tree-like mathematical model; the preset label classes are determined as child nodes in the tree-like mathematical model, and then the parent-child node relationship between the label classes is determined based on the affiliation between the label classes, such as Figure 2 The traffic participant object class and motion state in the are both part of the preset tag class, and the motion state belongs to the traffic participant object class. Therefore, in the tree mathematical model, the motion state will be the child node of the traffic participant object class, and the traffic participant object class will be the parent node of the motion state. Then, various tag types are classified and divided into corresponding tag classes. After a tag type is divided into a tag class, the tag type is determined as a leaf node under the tag class, such as Figure 2The time tag type in is classified into the general class, so the time tag type is determined as a leaf node under the general class, such as Figure 2 The motion speed tag type in Figure 2 v) is divided into the motion state label class, so the motion speed label type is determined as the leaf node under each motion state label class. Finally, after all label types are determined as corresponding leaf nodes, the basic tree-like mathematical model is constructed.

[0094] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, step S5 may include: constructing a first tree-like mathematical model according to the determined root node, child node, parent-child node relationship and leaf node; according to the driving scene type, presetting the evaluation weight of each first child node directly belonging to the root node in the first tree-like mathematical model to obtain a basic tree-like mathematical model.

[0095] In this embodiment, when evaluating the autonomous driving algorithm, the present application not only considers various evaluation indicators but also considers the complexity of the driving scene during the control process. Therefore, accurate autonomous driving performance evaluation results can be obtained for differentiated scenarios. However, the influencing factors of differentiated scenarios are different. Determining the driving scene complexity in the same way for each different driving scenario will also have a certain impact on the accuracy of the performance evaluation results. Therefore, when the present application calculates the driving scene complexity for different driving scenarios in the process of the autonomous driving algorithm controlling the vehicle's autonomous driving, the weights of each label class in the basic tree-like mathematical model are adjusted based on different driving scenarios to adapt to the calculation of the driving scene complexity of different driving scenarios, thereby improving the final autonomous driving performance evaluation results. For example, Figure 2 As shown in the figure, for urban roads, there will be more traffic lights and road congestion, so the impact of traffic on the complexity of driving scenes will be considered to a greater extent. Therefore, for urban road driving scenes, a larger weight is assigned to the traffic class; and for highways, there are basically no traffic lights and few traffic congestion. Therefore, for highway driving scenes, when a higher weight is assigned to the traffic class, the accuracy of the calculated driving scene complexity will be affected, which in turn affects the autonomous driving performance evaluation results. Therefore, for highway driving scenes, a smaller weight is assigned to the traffic class, and more weight is assigned to other label classes.

[0096] Specifically, the present application first constructs a first tree-like mathematical model based on the root node in the tree-like mathematical model determined by steps S01 to S04, the child nodes in the tree-like mathematical model, the parent-child node relationship between each child node in the tree-like mathematical model, and the leaf nodes in the tree-like mathematical model. Then, according to the driving scene type, the evaluation weights of each first child node directly belonging to the root node in the first tree-like mathematical model are preset to obtain a basic tree-like mathematical model. For example, Figure 2 As shown, when the driving scenario type is a highway driving scenario, the evaluation weight of the first subnode of the general class directly belonging to the root node in the first tree-like mathematical model is preset to be 0.3, the evaluation weight of the first subnode of the traffic class is 0.1, the evaluation weight of the first subnode of the road class is 0.3, and the evaluation weight of the first subnode of the traffic participant class is 0.3, so as to obtain the corresponding basic tree-like mathematical model; when the driving scenario type is an urban highway driving scenario, the evaluation weight of the first subnode of the general class directly belonging to the root node in the first tree-like mathematical model is preset to be 0.25, the evaluation weight of the first subnode of the traffic class is 0.25, the evaluation weight of the first subnode of the road class is 0.25, and the evaluation weight of the first subnode of the traffic participant class is 0.25, so as to obtain the corresponding basic tree-like mathematical model.

[0097] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, step S3 may include: determining the complexity values ​​corresponding to the label data of each node in the driving scene tree mathematical model according to the driving scene tree mathematical model and the target mapping relationship, wherein the target mapping relationship records the mapping relationship between different label data and different complexity values ​​of each node in the driving scene tree mathematical model; determining the complexity of the driving scene according to the complexity values ​​corresponding to the label data of all nodes in the determined driving scene tree mathematical model.

[0098] In this embodiment, the application first constructs a target mapping relationship, which records the mapping relationship between different label data and complexity of each leaf node (that is, each label type). For example, for weather label types, different label data include cloudy, dense fog, rain and snow, hail, typhoon, etc. The expert scoring sets the complexity value to 0.2 for cloudy, 0.6 for dense fog, 0.4 for rain and snow, 0.6 for hail, 0.9 for typhoon, etc.

[0099] In this embodiment, after obtaining the driving scene tree-like mathematical model, based on the specific label data recorded in each leaf node in the driving scene tree-like mathematical model, the complexity value of each leaf node in the driving scene tree-like mathematical model is determined from the target mapping relationship, and finally, based on the complexity values ​​of all leaf nodes in the driving scene tree-like mathematical model, a driving scene complexity is determined.

[0100] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, step S3 may include: determining the complexity values ​​corresponding to the label data of each node in the driving scene tree mathematical model according to the driving scene tree mathematical model and the target mapping relationship; determining the target complexity value of the node according to the complexity value corresponding to the label data of the node and the evaluation weight of the first child node to which the node belongs; determining the driving scene complexity according to the target complexity values ​​of all nodes in the determined driving scene tree mathematical model.

[0101] In this embodiment, when there are evaluation weights for each first child node directly belonging to the root node in the constructed basic tree-like mathematical model, an optional implementation of step S3 of the present application is: the present application first constructs a target mapping relationship, in which the mapping relationship between the different label data and complexity of each leaf node (that is, each label type) is recorded. After obtaining the driving scene tree-like mathematical model, the complexity value of each leaf node in the driving scene tree-like mathematical model is determined from the target mapping relationship based on the specific label data recorded in each leaf node in the driving scene tree-like mathematical model. For the complexity value of each leaf node determined, the complexity value of the leaf node is multiplied by the evaluation weight of the first node to which the leaf node belongs to obtain the target complexity value of the leaf node, and then the target complexity values ​​of all leaf nodes are added to obtain the driving scene complexity.

[0102] In this embodiment, in order to facilitate rapid scene complexity calculation, a simple scene complexity calculation method for drivers is proposed. The driving scene complexity of this method consists of two parts: road topology factor t and interactive target behavior complexity factor a. At this time, the driving scene complexity is expressed as: C = t + a, where the road topology factor is shown in Table 1 below, and the interactive target behavior complexity factor is shown in Table 2 below:

[0103] Table 1 Table 2

[0104]

[0105] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, step S5 may include: obtaining an autonomous driving performance score by bringing the values ​​of various evaluation indicators and the complexity of the driving scene into a driving performance evaluation algorithm for calculation. The driving performance evaluation algorithm expression is: S = C (ω1I1 +

[0106] ω2I2+...+ω n I n ), where S is the autonomous driving performance score, C is the complexity of the driving scene, and ω n is the weight of the nth evaluation index, I n is the value of the nth evaluation index.

[0107] In this embodiment, due to different evaluation requirements of the autonomous driving algorithm, this application sets corresponding weights for each evaluation indicator when determining the autonomous driving performance score based on various evaluation indicators, so that the evaluation strategy can be adjusted according to the different evaluation requirements of the autonomous driving algorithm. For example, when evaluating the autonomous driving performance of the autonomous driving algorithm, the demand is to pay more attention to the anthropomorphic performance of the autonomous driving algorithm, so a larger weight can be set for the anthropomorphic evaluation indicator, so that the final autonomous driving performance score can better reflect the anthropomorphism; the demand is to pay more attention to the safety performance of the autonomous driving algorithm, so a larger weight can be set for the safety evaluation indicator, so that the final autonomous driving performance score can better reflect safety.

[0108] Specifically, the values ​​of various evaluation indicators and the complexity of the driving scene determined by steps S1 to S4 are brought into the driving performance evaluation algorithm for calculation to obtain the automatic driving performance score. The driving performance evaluation algorithm expression is: S = C (ω1I1+ω2I2+...+ω b I n ), where S is the autonomous driving performance score, C is the complexity of the driving scene, and ω n is the weight of the nth evaluation index, I n is the value of the nth evaluation index. The weights of various evaluation indicators can be set according to the evaluation requirements, and the sum of the weights of all types of evaluation indicators is 1. When the evaluation indicators include safety evaluation indicators, efficiency evaluation indicators, comfort evaluation indicators and anthropomorphic evaluation indicators, the value of n is 4.

[0109] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, when the evaluation index is a safety evaluation index, the value of the evaluation index is determined according to the driving scene data, including: the value of the safety evaluation index is obtained by bringing the relative distance and respective speeds between the self-vehicle and the target vehicle in the driving scene data into the safety evaluation algorithm for calculation, and the safety evaluation algorithm expression is:

[0110]

[0111] Among them, U is the safety evaluation index, U x is the vertical risk, U y is the horizontal risk, k lon is the virtual longitudinal spring stiffness, x is the relative distance between the ego vehicle and the target vehicle, x0 is the expected relative distance between the ego vehicle and the target vehicle, c lon is the longitudinal virtual damping coefficient, v p is the target vehicle speed, v x is the vehicle speed, c lat is the lateral virtual damping coefficient, v y is the lateral speed of the vehicle, y is the lateral offset of the vehicle from the center line of the current lane, k lat is the virtual lateral spring stiffness, U x,max To customize the maximum vertical value and U y,max It is the custom horizontal maximum value.

[0112] In this embodiment, considering that the risk field can reflect the potential danger level in the current environment of the ego vehicle, this application introduces it to evaluate the safety of the scene. In order to make the designed risk close to the risk felt by human drivers, a physical model of virtual springs and damping is introduced to simulate the force between vehicles. Specifically: when the evaluation index is a safety evaluation index, the relative distance and respective speeds between the ego vehicle and the target vehicle in the driving scene data are brought into the safety evaluation algorithm for calculation to obtain the value of the safety evaluation index. The safety evaluation algorithm expression is: Among them, U is the safety evaluation index, U x is the vertical risk, U y is the horizontal risk, k lon is the virtual longitudinal spring stiffness, which is a known quantity set in advance, x is the relative distance between the ego vehicle and the target vehicle, x0 is the expected relative distance between the ego vehicle and the target vehicle, c lon is the longitudinal virtual damping coefficient, which is a known quantity set in advance, v p is the target vehicle speed, vx is the vehicle speed, c lat is the lateral virtual damping coefficient, which is a known quantity set in advance, v y is the lateral speed of the vehicle, y is the lateral offset of the vehicle from the center line of the current lane, k lat is the virtual lateral spring stiffness, which is a known quantity set in advance, U x,max To customize the maximum vertical value and U y,max It is the custom horizontal maximum value.

[0113] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, when the evaluation index is an efficiency evaluation index, the value of the evaluation index is determined according to the driving scene data, including: the value of the efficiency evaluation index is obtained by bringing the maximum speed limit of the lane and the vehicle speed in the driving scene data into the efficiency evaluation algorithm for calculation, and the efficiency evaluation algorithm expression is: Among them, I e is the efficiency evaluation index, v x is the vehicle speed, v ego,x The maximum speed limit for the lane.

[0114] In this embodiment, efficiency can be directly measured by the length of time, but different traffic flow scenarios will make the time consumption different when the starting point and the end point are specified. However, the higher the speed of the vehicle, the closer it is to the maximum speed of the vehicle in front of the vehicle, and the higher the driving efficiency of the vehicle. Therefore, when the evaluation index is the efficiency evaluation index, the maximum speed limit of the lane and the speed of the vehicle in the driving scene data are brought into the efficiency evaluation algorithm for calculation to obtain the value of the efficiency evaluation index. The expression of the efficiency evaluation algorithm is: Among them, I e is the efficiency evaluation index, v x is the vehicle speed, v ego,x The maximum speed limit for the lane.

[0115] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, when the evaluation index is a comfort evaluation index, the value of the evaluation index is determined according to the driving scene data, including: obtaining the value of the comfort evaluation index by bringing the acceleration of the vehicle in the driving scene data into the comfort evaluation algorithm for calculation, and the expression of the comfort evaluation algorithm is: Among them, I c is the comfort evaluation index, a index is the vehicle acceleration index, a lonis the longitudinal acceleration, a lat is the lateral acceleration, t is the sampling frequency, i is the i-th sampling, N is the number of data points sampled per unit time, jerk index is the acceleration index, jerk lon is the longitudinal acceleration, jerk lat is the lateral acceleration, is the value of the vehicle acceleration, is the human body’s sensitive acceleration value, is the value of the vehicle's acceleration, It is the sensitive acceleration value of human body.

[0117] In this embodiment, the present application adopts the vehicle acceleration index and the jerk index as the comfort evaluation index. Specifically: when the evaluation index is the comfort evaluation index, the vehicle acceleration in the driving scene data is brought into the comfort evaluation algorithm for calculation to obtain the value of the comfort evaluation index. The comfort evaluation algorithm expression is: Among them, I c is the comfort evaluation index, a index is the vehicle acceleration index, a lon is the longitudinal acceleration, a lat is the lateral acceleration, t is the sampling frequency, preferably 25, i is the i-th sampling, N is the number of data points sampled per unit time, jerk index is the acceleration index, jerk lon is the longitudinal acceleration, jerk lat is the lateral acceleration, is the value of the vehicle acceleration, is the human body's sensitive acceleration value, preferably 2.8. is the value of the vehicle's acceleration, is the human body's sensitive acceleration value, and the preferred value is 1.14.

[0119] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a performance evaluation method for autonomous driving. In the performance evaluation method for autonomous driving, when the evaluation index is an anthropomorphic evaluation index, the value of the evaluation index is determined according to the driving scene data, including: the value of the anthropomorphic evaluation index is obtained by bringing the vehicle trajectory in the driving scene data into the anthropomorphic evaluation algorithm for calculation, and the expression of the anthropomorphic evaluation algorithm is:

[0120] D(i,j)=(x i +y i ) 2+min([D(i-1,j),D(i,j-1),D(i-1,j-1)]);

[0121] Among them, the actual trajectory sequence X={x1,x2,x3,…,x n}, x i is the actual coordinate position of the vehicle at the i-th moment, and the trajectory sequence to be evaluated Y = {y1, y2, y3, …, y m},y i is the coordinate position of the vehicle to be evaluated at the i-th moment, and n and m represent the lengths of the two sequences respectively.

[0122] In this embodiment, when the evaluation index is an anthropomorphic evaluation index, the present application obtains the value of the anthropomorphic evaluation index by bringing the vehicle trajectory in the driving scene data into the anthropomorphic evaluation algorithm for calculation. The expression of the anthropomorphic evaluation algorithm is: D(i,j)=(x i +y i ) 2 +

[0123] min([D(i-1,j),D(i,j-1),D(i-1,j-1)]); where the actual trajectory sequence X = {x1,x2,x3,…,x n}, x i is the actual coordinate position of the vehicle at the i-th moment, and the trajectory sequence to be evaluated Y = {y1, y2, y3, …, y m},y i is the coordinate position of the vehicle to be evaluated at the i-th moment, and n and m represent the lengths of the two sequences respectively.

[0124] Based on the same inventive concept, an embodiment of the present application provides a performance evaluation system for autonomous driving, such as Figure 4 As shown, the system 400 includes:

[0125] The tag data determination module 401 is used to determine the tag data of various tag types in the driving scene data according to various preset tag types;

[0126] A driving scene tree mathematical model construction module 402 is used to construct a driving scene tree mathematical model according to label data of various label types in the driving scene data and a basic tree mathematical model, wherein each leaf node in the basic tree mathematical model is composed of various preset label types;

[0127] A driving scene complexity determination module 403, used to determine the driving scene complexity according to a tree-like mathematical model of the driving scene;

[0128] An evaluation index value determination module 404 is used to determine the value of the evaluation index according to the driving scene data;

[0129] The automatic driving performance score determination module 405 is used to determine the automatic driving performance score according to the values ​​of various evaluation indicators and the complexity of the driving scene.

[0130] Based on the same inventive concept, an embodiment of the present application provides an electronic device, including: a processor, a memory, and a computer program stored in the memory and running on the processor. When the computer program is executed by the processor, it implements the steps in a performance evaluation method for autonomous driving as described in the first aspect of the present application.

[0131] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in a performance evaluation method for autonomous driving as described in the first aspect of the present application are implemented.

[0132] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0133] It should be noted that, for the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0134] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0135] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0136] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0137] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0139] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0140] Finally, it should be noted that, in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements. The above is a detailed introduction to a performance evaluation method, system and product of an autonomous driving provided by the present application. The present application is explained in detail using specific examples. The description of the above embodiments is only used to help understand the method and its core idea of ​​the present application; at the same time, for those skilled in the art, according to the idea of ​​the present application, there will be changes in the specific implementation and scope of application. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A performance evaluation method for autonomous driving, characterized in that: The method comprises: Determine label data of various label types in the driving scene data according to various preset label types; According to the label data of various label types in the driving scene data and the basic tree-like mathematical model, a driving scene tree-like mathematical model is constructed, wherein each leaf node in the basic tree-like mathematical model is composed of various preset label types; Determine the complexity of the driving scenario based on the tree-like mathematical model of the driving scenario; Determine the value of the evaluation index based on the driving scenario data; The autonomous driving performance score is determined based on the values ​​of various evaluation indicators and the complexity of the driving scenario.

2. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: According to the label data of various label types in the driving scene data and the basic tree-like mathematical model, a driving scene tree-like mathematical model is constructed, including: According to various label types in the driving scene data, determining the node positions of various label types in the basic tree mathematical model; The label data of various label types in the driving scene data are placed into their corresponding node positions to obtain a tree-like mathematical model of the driving scene.

3. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: Construct a basic tree-like mathematical model, including: Determine the natural driving scene node as the root node of the tree mathematical model; Determine each preset label class as a child node of the tree-like mathematical model; According to the affiliation between the tag classes, the parent-child node relationship between the tag classes is determined; Identify various tag types as leaf nodes under the tag class; According to the determined root nodes, child nodes, parent-child node relationships and leaf nodes, a basic tree-like mathematical model is constructed.

4. The method for evaluating the performance of an autonomous driving system according to claim 3, characterized in that: According to the determined root nodes, child nodes, parent-child node relationships and leaf nodes, a basic tree-like mathematical model is constructed, including: Constructing a first tree-like mathematical model according to the determined root node, child nodes, parent-child node relationships, and leaf nodes; According to the driving scenario type, the evaluation weights of the first child nodes directly belonging to the root node in the first tree-like mathematical model are preset to obtain a basic tree-like mathematical model.

5. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: According to the tree-like mathematical model of the driving scenario, the complexity of the driving scenario is determined, including: Determine the complexity values ​​corresponding to the label data of each node in the driving scene tree-like mathematical model according to the driving scene tree-like mathematical model and the target mapping relationship, wherein the target mapping relationship records the mapping relationship between different label data of each node in the driving scene tree-like mathematical model and different complexity values; The complexity of the driving scenario is determined according to the complexity values ​​corresponding to the label data of all nodes in the determined tree-like mathematical model of the driving scenario.

6. The method for evaluating the performance of an autonomous driving system according to claim 4, characterized in that: According to the tree-like mathematical model of the driving scenario, the complexity of the driving scenario is determined, including: Determine the complexity value corresponding to the label data of each node in the driving scene tree mathematical model according to the driving scene tree mathematical model and the target mapping relationship; determine the target complexity value of the node according to the complexity value corresponding to the label data of the node and the evaluation weight of the first child node to which the node belongs; The complexity of the driving scenario is determined according to the target complexity values ​​of all nodes in the determined driving scenario tree mathematical model.

7. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: The autonomous driving performance score is determined based on the values ​​of various evaluation indicators and the complexity of the driving scenario, including: The automatic driving performance score is obtained by bringing the values ​​of various evaluation indicators and the complexity of the driving scene into the driving performance evaluation algorithm for calculation. The expression of the driving performance evaluation algorithm is: S = C (ω1I1+ω2I2+...+ω n I n ); Among them, S is the autonomous driving performance score, C is the complexity of the driving scene, ω n is the weight of the nth evaluation index, I n is the value of the nth evaluation index.

8. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: When the evaluation index is a safety evaluation index, the value of the evaluation index is determined according to the driving scenario data, including: The relative distance between the self-vehicle and the target vehicle in the driving scene data and their respective speeds are brought into the safety assessment algorithm for calculation to obtain the value of the safety evaluation index. The safety assessment algorithm expression is: Among them, U is the safety evaluation index, U x is the vertical risk, U y is the horizontal risk, k lon is the virtual longitudinal spring stiffness, x is the relative distance between the ego vehicle and the target vehicle, x0 is the expected relative distance between the ego vehicle and the target vehicle, c lon is the longitudinal virtual damping coefficient, v p is the target vehicle speed, v x is the vehicle speed, c lat is the lateral virtual damping coefficient, v y is the lateral speed of the vehicle, y is the lateral offset of the vehicle from the center line of the current lane, k lat is the virtual lateral spring stiffness, U x,max To customize the maximum vertical value and U y,max It is the custom horizontal maximum value.

9. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: When the evaluation index is an efficiency evaluation index, the value of the evaluation index is determined according to the driving scenario data, including: The value of the efficiency evaluation index is obtained by bringing the lane maximum speed limit and the vehicle speed in the driving scenario data into the efficiency evaluation algorithm for calculation. The efficiency evaluation algorithm expression is: Among them, I e is the efficiency evaluation index, v x is the vehicle speed, v ego,,x The maximum speed limit for the lane.

10. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: When the evaluation index is a comfort evaluation index, the value of the evaluation index is determined according to the driving scenario data, including: The value of the comfort evaluation index is obtained by bringing the vehicle acceleration in the driving scene data into the comfort evaluation algorithm for calculation. The expression of the comfort evaluation algorithm is: Among them, I c is the comfort evaluation index, a index is the vehicle acceleration index, a lon is the longitudinal acceleration, a lat is the lateral acceleration, t is the sampling frequency, i is the i-th sampling, N is the number of data points sampled per unit time, jerk index is the acceleration index, jerk lon is the longitudinal acceleration, jerk lat is the lateral acceleration, is the value of the vehicle acceleration, is the human body’s sensitive acceleration value, is the value of the vehicle's acceleration, It is the sensitive acceleration value of human body.

11. The method for evaluating the performance of an autonomous driving system according to claim 1, characterized in that: When the evaluation index is an anthropomorphic evaluation index, the value of the evaluation index is determined according to the driving scene data, including: The value of the anthropomorphic evaluation index is obtained by bringing the vehicle trajectory in the driving scene data into the anthropomorphic evaluation algorithm for calculation. The expression of the anthropomorphic evaluation algorithm is: Among them, the actual trajectory sequence X = {x t ,x2,x3,…,x n }, x i is the actual coordinate position of the vehicle at the i-th moment, and the trajectory sequence to be evaluated Y = {y1, y2, y3, …, y m },y i is the coordinate position of the vehicle to be evaluated at the i-th moment, and n and m represent the lengths of the two sequences respectively.

12. A performance evaluation system for autonomous driving, characterized in that: The system comprises: a label data determination module, for determining label data of various label types in driving scene data according to various preset label types; A driving scene tree mathematical model construction module is used to construct a driving scene tree mathematical model according to label data of various label types in the driving scene data and a basic tree mathematical model, wherein each leaf node in the basic tree mathematical model is composed of various preset label types; a driving scene complexity determination module is used to determine the driving scene complexity according to the driving scene tree mathematical model; An evaluation index value determination module is used to determine the value of the evaluation index according to the driving scene data; The autonomous driving performance score determination module is used to determine the autonomous driving performance score based on the values ​​of various evaluation indicators and the complexity of the driving scenario.

13. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and running on the processor, wherein the computer program, when executed by the processor, implements the steps in a performance evaluation method for autonomous driving as described in any one of claims 1 to 11.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in a performance evaluation method for autonomous driving as described in any one of claims 1 to 11.

Images