A continuous scene testing method, device and medium based on virtual map

Abstract

The present invention relates to a method, device and medium for continuous scene testing based on a virtual map, the method comprising the following steps: according to the characteristic information of the actual road network, a virtual map is established, a test environment is built, and multiple test scenes are set, and a test vehicle is connected to the test environment; vehicle status and environmental information are obtained in real time, and according to the vehicle status and environmental information, the optimal test scene is matched by a preset matching rule, and the optimal test scene is loaded into the test environment for scene testing. After the scene test is completed, this step is re-executed until the continuity test of multiple test scenes is completed. Compared with the prior art, the present invention realizes continuous testing on a virtual map, and there is a certain linkage relationship between the scenes, so that the scene test is more authentic; at the same time, each scene selection has a certain randomness, which can better test the algorithm under test and improve the effectiveness of the test.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving vehicle testing, and in particular to a continuous scene testing method, equipment, and medium based on a virtual map. Background Art

[0002] At present, the simulation test scenarios of autonomous driving are basically built according to the requirements of the test scenarios. Some fragmented test scenarios are built for single or limited test requirements. Different test requirements can only be tested separately through different fragmented test scenarios, and each fragmented test scenario has basically no connection with each other. This fragmented simple test scenario cannot achieve the linkage response between different scenarios in the real world, and cannot reflect the coupling relationship between real scenarios. In order to solve the above problems, a test method for continuous scenes based on a large virtual map is designed, which is very meaningful for the current field of autonomous driving test technology.

[0003] Chinese patent application number CN201910048241.6 discloses a vehicle assisted driving test system and method, the method comprising the following steps: obtaining the performance parameters of the tested vehicle; generating VR scene information; running the vehicle simulation model of the tested vehicle; controlling the operating parameters of the vehicle simulation model based on the performance parameters of the tested vehicle and the VR scene information; and changing the VR scene information based on the operating parameters of the vehicle simulation model. The test of vehicle assisted driving is realized. However, there is a lack of connection between the various test scenarios of this method, the test scenarios are separated from the actual environment, and the test effect is poor.

[0004] In summary, the current testing methods have the following shortcomings:

[0005] (1) Each scenario is independent of each other and lacks linkage between scenarios, which makes the scenario test lack authenticity and is quite different from the actual driving scenarios in real life.

[0006] (2) Each scene selection lacks randomness, and the initial state of each traffic participant also lacks a certain degree of randomness. The lack of randomness will lead to poor test results of the tested algorithm and low test effectiveness. Summary of the invention

[0007] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and provide a continuous scene testing method based on a virtual map, which can realize the linkage response between different scenes in the real world and reflect the coupling relationship between real scenes.

[0008] The purpose of the present invention can be achieved by the following technical solutions:

[0009] One aspect of the present invention provides a continuous scenario testing method based on a virtual map, comprising the following steps: establishing a virtual map and a test environment according to actual road network feature information, and setting a plurality of test scenarios that are mutually related through trigger conditions and completion conditions, and connecting a test vehicle to the test environment; acquiring vehicle status and environmental information in real time, matching a plurality of candidate test scenarios according to preset matching rules based on the vehicle status and environmental information, randomly selecting a target test scenario, loading the target test scenario into the test environment for scenario testing, and re-executing this step after the preset scenario completion conditions are met, until the continuity test of the plurality of test scenarios is completed.

[0010] As a preferred technical solution, the road network characteristic information includes road structure information and / or traffic facility information.

[0011] As a preferred technical solution, the test scenario includes a standard test scenario and / or an accident hazard scenario.

[0012] As a preferred technical solution, the process of obtaining the preset matching rules includes the following steps: obtaining the matching rules based on the first trigger condition information and the second trigger condition information, wherein the process of obtaining the first trigger condition information and the second trigger condition information includes the following steps: obtaining the road structure information and the motion state information of the vehicle in the test scene based on the element composition information in each test scene; obtaining the first trigger condition information in the test scene based on the road structure information, and dividing the multiple test scenes into multiple test scene groups; for each of the test scene groups, obtaining the second trigger condition information in the test scene based on the motion state information of the vehicle in the test scene.

[0013] As a preferred technical solution, the element composition information includes one or more of road structure information, traffic facility information, natural environment information, and traffic participant information.

[0014] As a preferred technical solution, the natural environment information includes weather information and / or time information and / or light information.

[0015] As a preferred technical solution, the traffic participant information includes one or more of vehicle type information, relative position information with respect to the vehicle under test, relative speed information with respect to the vehicle under test, and acceleration information of the vehicle itself.

[0016] As a preferred technical solution, the process of loading the optimal test scenario into the test environment for scenario testing includes the following steps:

[0017] Obtain the element composition information of the test scenario under each test scenario, and obtain the selection range of each variable; generate a matching environment and traffic participants according to the natural environment information and traffic participant information, select an initial value in the selection range of each variable, and complete the initialization; decompose the complete optimal test scenario into multiple scene segments, and control the test vehicle to complete the above scene segments in sequence according to the vehicle status information obtained in real time, until the test vehicle reaches the end point of the current test scenario, and completes the test under the current test scenario.

[0018] Another aspect of the present invention provides an electronic device including one or more processors and a memory, wherein the memory stores one or more programs, and the one or more programs include instructions for executing the above-mentioned virtual map-based continuous scenario testing method.

[0019] Another aspect of the present invention provides a computer-readable storage medium comprising one or more programs for execution by one or more processors of an electronic device, wherein the one or more programs include instructions for executing the above-mentioned virtual map-based continuous scenario testing method.

[0020] Compared with the prior art, the present invention has the following advantages:

[0021] (1) Continuous testing is performed on a virtual map. The completion conditions of the previous test scenario will affect the triggering conditions of the next test scenario, so that there is a certain linkage relationship between the scenarios. The various test scenarios are related to each other, making the scenario test more realistic and closer to the real driving scenarios in real life, so that the test results are closer to reality.

[0022] (2) Each scene selection has a certain degree of randomness, and the initial state of each traffic participant also has a certain degree of randomness, which increases the randomness of the scene. Such randomness can better test the algorithm under test and improve the effectiveness of the test. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 Flow chart of a continuous scene testing method based on a virtual map in an embodiment. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] Example 1

[0026] like Figure 1 The flowchart of the continuous scene testing method based on the virtual map provided in this embodiment includes the following steps:

[0027] Step S1, based on the actual road network feature information, a virtual map is established, a test environment is built, and multiple test scenarios are set up, and a test vehicle is connected to the test environment.

[0028] Step S2, real-time acquisition of vehicle status and environmental information, matching the optimal test scenario according to the vehicle status and environmental information through preset matching rules, loading the optimal test scenario into the test environment for scenario testing, and after the scenario testing is completed, re-execute this step until the continuity test of multiple test scenarios is completed.

[0029] In step S1, the road network feature information includes road structure information and traffic facility information. Road structure information (such as straight roads, curves, ramps, elevated roads, tunnels, intersections, single lanes, multiple lanes, lane line types (solid lines, dashed lines, etc.), lane types), traffic facility information (such as traffic lights, warning signs, zebra crossings, etc.) and other real-world traffic environments and elements. Test scenarios include standard test scenarios and accident hazard scenarios. Standard test scenarios include test scenarios required by regulatory standards such as ACC, AEB, and LKA. Accident hazard scenarios include common accident hazard scenarios, such as ghost head-on scenarios, backlight scenarios, and other car cut-in and cut-out scenarios. Specific test scenarios will be continuously expanded according to actual test needs and special scenarios collected from the real world, forming a continuously updated and expanded test scenario list.

[0030] In step S2, the process of obtaining the preset matching rule includes the following steps:

[0031] According to the first trigger condition information and the second trigger condition information, a matching rule is obtained,

[0032] The process of obtaining the first trigger condition information and the second trigger condition information includes the following steps:

[0033] According to the element composition information in each test scenario, the road structure information and the motion state information of the vehicle in the test scenario are obtained;

[0034] According to the road structure information (such as whether it is a straight road, whether it is multi-lane, whether it is a ramp, etc.), the first trigger condition information in the test scenario is obtained, and the multiple test scenarios are divided into multiple test scenario groups;

[0035] For each test scenario group, the second trigger condition information (such as the speed, acceleration, lane change, etc. of the vehicle being driven) in the test scenario is obtained according to the motion state information of the vehicle in the test scenario.

[0036] The element composition information includes road structure information, traffic facility information, natural environment information (daytime, evening, night, backlight, glare, rain, snow, fog, etc.), traffic participant information (traffic participant types: cars, trucks, vans, bicycles, electric vehicles, people, etc., relative position with the vehicle under test, relative speed, acceleration requirements of the vehicle itself, etc. The relative position, relative speed, acceleration information, etc. are all range values, and a state value will be randomly selected within these ranges during subsequent initialization).

[0037] The process of loading the optimal test scenario into the test environment for scenario testing includes the following steps:

[0038] Obtain the element composition information of each test scenario and the selection range of each variable;

[0039] Generate matching environment and traffic participants according to natural environment information and traffic participant information, select initial values ​​in the selection range of each variable, and complete initialization;

[0040] The complete optimal test scenario is decomposed into multiple scenario segments. According to the real-time vehicle status information, the test vehicle is controlled to complete the above scenario segments in sequence until the test vehicle reaches the end point of the current test scenario and completes the test under the current test scenario.

[0041] The specific testing method adopts the method of building a dynamic test scene based on background vehicle path control in Chinese patent application number CN202210708522.1 for testing.

[0042] The core idea of ​​this method is to use a virtual map with multiple real road network features and surrounding environments, and design test scenarios targeting the road structure characteristics according to the different road structures and environmental characteristics of the map. Then, after the test vehicle is connected to the test environment, the movement state of the test vehicle and the surrounding road environment are monitored in real time. According to this information, a suitable test scenario is selected, and the test scenario is dynamically generated according to the real-time state of the vehicle. After the vehicle completes the corresponding scenario, the vehicle state is monitored again. When the next surrounding environment that meets the test scenario conditions is found, the test scenario is dynamically generated again according to the real-time state of the vehicle. In this way, multiple test scenarios can be continuously generated according to the real-time state of the vehicle in a larger virtual map, realizing the linkage between test scenarios, which is more in line with the vehicle's operating environment in the real world.

[0043] The above-mentioned real road network characteristics and surrounding environment information include: road structure information (such as straight roads, curves, ramps, elevated roads, tunnels, intersections, single lanes, multiple lanes, lane line types (solid lines, dashed lines, etc.), lane types), traffic facility information (such as traffic lights, warning signs, zebra crossings, etc.) and other real-world traffic environments and elements;

[0044] The above-mentioned test scenarios include test scenarios required by regulatory standards such as ACC, AEB, and LKA, as well as common accident risk scenarios such as ghost peeking scenarios, backlight scenarios, and other car cut-in and cut-out scenarios. The specific test scenarios will be continuously expanded according to actual test needs and special scenarios collected from the real world, forming a continuously updated and expanded test scenario list;

[0045] Analyze the constituent elements of each test scenario in the above scenario list, including road structure information, traffic facility information, natural environment information, traffic participant information, etc.;

[0046] For the above-mentioned scene constituent elements, the trigger conditions of each scene are analyzed, such as the road structure conditions (such as: whether it is a straight road, whether it is multi-lane, whether it is a ramp, etc.). The vehicle motion state scene classifies the scene list according to the trigger conditions of the road structure. On the basis of this classification, the motion state required by the tested vehicle in each scene is further analyzed as another trigger condition (such as the speed, acceleration, lane change, etc. of the tested vehicle). The specific trigger conditions can be designed by yourself.

[0047] The specific steps are as follows:

[0048] 1. During the test, the motion state of the vehicle under test and the structural characteristics of the road on which the vehicle under test is located, etc., which match the above trigger conditions, are obtained in real time. Based on this information, a suitable test scenario is determined and a scenario is randomly selected from the suitable test scenarios as the scenario for this test;

[0049] 2. According to the test scenarios determined above, determine the required natural environment information (daytime, evening, night, backlight, glare, rain, snow, fog, etc.) and traffic participant information (traffic participant types: cars, trucks, vans, bicycles, electric vehicles, people, etc., relative position and relative speed with the vehicle under test, acceleration requirements of the vehicle itself, etc., relative position, relative speed, acceleration information, etc. are all range values, and a state value will be randomly selected within these ranges during subsequent initialization);

[0050] According to the above-analyzed natural environment information and traffic participant information, the corresponding environment and traffic participants are generated in real time in the simulation simulator to realize the initialization of the scene;

[0051] 3. After the traffic participants are generated, the method in the previously written patent of an efficient method for building a dynamic test scene based on background vehicle path control is used to complete the control of the background vehicles in the entire scene and complete the test of this scene. During this period, the scene will not be reselected until the current scene is completed;

[0052] After the above test scene is completed, proceed to steps 1, 2, and 3 to open another test scene, and so on in sequence to achieve continuous testing of multiple scenes on a large map;

[0053] The above methods can be briefly summarized as follows:

[0054] First, a virtual map is generated, which contains relevant characteristics of the road environment in the real world and is built according to the test requirements. It can also be built directly based on an area in the real world as a template.

[0055] Then a scenario list is generated, which contains many scenarios and all the element information related to the scenario test. The scenario list is continuously expanded and updated;

[0056] Next, according to the state information of the tested vehicle and the surrounding environment information obtained from the simulator in real time during the test, it is determined which scenario is suitable at this moment. The specific judgment process is designed based on the behavior tree. According to the trigger conditions of different scenarios, it is determined whether the trigger of this type of scenario is met at this moment, and finally a scenario is randomly selected from the scenarios that meet the trigger conditions as the test scenario of this section;

[0057] Complete initialization in the simulator according to the extracted scenario, including initialization of the environment and initialization of traffic participants;

[0058] After the initialization is completed, the method in the previously written patent of the efficient construction method of dynamic test scene based on background vehicle path control is used to complete the control of the background vehicle in the entire scene and complete the test of this scene without changing the scene during the process;

[0059] This process is repeated to complete continuous simulation tests of multiple scenarios based on virtual maps.

[0060] The behavior tree contains the trigger conditions for each scene to start and the completion conditions for the scene to complete. If you are not currently in a test scene, you will determine which scenes can be triggered based on the trigger conditions. Randomly select a scene from these scenes as the test scene for this environment. After entering the test scene, it will continuously detect whether the completion conditions of the scene are met to determine whether the scene is completed. If the scene is not completed, it will not determine whether the trigger conditions of the scene are met. When the scene is not triggered, search for the trigger conditions, select the scene, generate the scene, and determine whether the scene is completed after entering the scene. During this period, the trigger conditions will not be searched. After the scene is completed, the trigger conditions will be searched again, and the cycle will continue.

[0061] This method realizes the function of continuous testing on a virtual map. The previous scene will affect the next scene, and there is a certain linkage between the scenes, which makes the scene test more realistic and closer to the real driving scene in real life. At the same time, each scene selection has a certain randomness, and the initial state of each traffic participant also has a certain randomness, which increases the randomness of the scene. Such randomness can better test the algorithm under test and improve the effectiveness of the test.

[0062] Example 2

[0063] This embodiment provides an electronic device, including one or more processors and a memory, wherein a plurality of programs are stored in the memory, and the plurality of programs include instructions for executing the continuous scene testing method based on a virtual map as described in Embodiment 1.

[0064] Example 3

[0065] This embodiment provides a computer-readable storage medium, including multiple programs for execution by one or more processors of an electronic device, and the multiple programs include instructions for executing the continuous scenario testing method based on a virtual map as described in Example 1.

[0066] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A continuous scene testing method based on a virtual map, characterized in that: The steps include: According to the actual road network characteristic information, a virtual map is established, a test environment is built, and multiple test scenarios are set up which are interconnected by trigger conditions and completion conditions, and a test vehicle is connected to the test environment; Acquire vehicle status and environment information in real time, match multiple candidate test scenarios according to the vehicle status and environment information through preset matching rules, randomly select a target test scenario, load the target test scenario into the test environment for scenario testing, and re-execute this step after meeting preset scenario completion conditions until the continuity test of multiple test scenarios is completed. The process of obtaining the preset matching rules includes the following steps: According to the first trigger condition information and the second trigger condition information, the matching rule is acquired, The process of obtaining the first trigger condition information and the second trigger condition information includes the following steps: According to the element composition information in each test scenario, the road structure information and the motion state information of the vehicle in the test scenario are obtained; According to the road structure information, first trigger condition information in the test scenario is acquired, and the multiple test scenarios are divided into multiple test scenario groups; For each of the test scenario groups, second trigger condition information in the test scenario is obtained according to the motion state information of the vehicle in the test scenario.

2. A virtual map-based continuous scene testing method according to claim 1, characterized in that: The road network characteristic information includes road structure information and / or traffic facility information.

3. The method for continuous scene testing based on a virtual map according to claim 1, characterized in that: The test scenarios include standard test scenarios and / or accident hazard scenarios.

4. The method for continuous scene testing based on a virtual map according to claim 1, characterized in that: The element composition information includes one or more of road structure information, traffic facility information, natural environment information, and traffic participant information.

5. The method for continuous scene testing based on a virtual map according to claim 4, characterized in that: The natural environment information includes weather information and / or time information and / or light information.

6. The method for continuous scene testing based on a virtual map according to claim 4, characterized in that: The traffic participant information includes one or more of vehicle type information, relative position information with respect to the detected vehicle, relative speed information with respect to the detected vehicle, and acceleration information of the vehicle itself.

7. The method for continuous scene testing based on a virtual map according to claim 1, characterized in that: The process of loading the optimal test scenario into the test environment for scenario testing includes the following steps: Obtain the element composition information of each test scenario and the selection range of each variable; Generate matching environment and traffic participants according to natural environment information and traffic participant information, select initial values ​​within the selection range of each variable, and complete initialization; The complete optimal test scenario is decomposed into multiple scenario segments. According to the real-time vehicle status information, the test vehicle is controlled to complete the above scenario segments in sequence until the test vehicle reaches the end point of the current test scenario and completes the test under the current test scenario.

8. An electronic device, characterized in that: It comprises one or more processors and a memory, wherein one or more programs are stored in the memory, and the one or more programs include instructions for executing the continuous scene testing method based on a virtual map as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that: It includes one or more programs for execution by one or more processors of an electronic device, and the one or more programs include instructions for executing the continuous scene testing method based on a virtual map as described in any one of claims 1-7.

Images