CN-115935620-B - Simulation test scene recommendation method and device, storage medium and computer equipment

CN115935620BCN 115935620 BCN115935620 BCN 115935620BCN-115935620-B

Abstract

The application provides a simulation test scene recommending method, a device, a storage medium and computer equipment, wherein the method comprises the steps of acquiring first automatic driving data output by a first automatic driving algorithm under each simulation test scene and second automatic driving data output by a second automatic driving algorithm under each simulation test scene; the method comprises the steps of determining a difference value of each simulation test scene under each difference evaluation index according to first automatic driving data, second automatic driving data and preset each difference evaluation index, screening simulation test scenes with large difference degree from each simulation test scene according to each difference value corresponding to each difference evaluation index as target scenes corresponding to each difference evaluation index aiming at each difference evaluation index, determining a recommended scene set according to target scenes corresponding to each difference evaluation index, wherein the recommended scene set comprises at least one target scene. The application can improve the test efficiency.

Inventors

WU JIACHEN
ZHENG ZIWEI
TAN WEIHUA

Assignees

广州文远知行科技有限公司

Dates

Publication Date: 20260508
Application Date: 20221123

Claims (9)

1. A simulation test scene recommendation method, the method comprising: acquiring first automatic driving data output by a first automatic driving algorithm under each simulation test scene and second automatic driving data output by a second automatic driving algorithm under each simulation test scene; Determining a difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index; Aiming at each difference evaluation index, screening out a simulation test scene with large difference degree from the simulation test scenes according to each difference value corresponding to the difference evaluation index as a target scene corresponding to the difference evaluation index; determining a recommended scene set according to target scenes corresponding to the difference evaluation indexes, wherein the recommended scene set comprises at least one target scene, and the method specifically comprises the following steps: Accumulating the number of the target scenes corresponding to the difference evaluation indexes to obtain the total number of the target scenes; respectively determining the target scene duty ratio corresponding to each difference evaluation index according to the total number of the target scenes and the number of the target scenes corresponding to each difference evaluation index; acquiring the total number of preset recommended scenes; selecting K recommended scenes from the target scenes corresponding to the difference evaluation indexes according to the total number of the recommended scenes and the target scene occupation ratio corresponding to the difference evaluation indexes to obtain a recommended scene set comprising K recommended scenes, wherein K is the total number of the recommended scenes.
2. The simulation test scene recommendation method according to claim 1, wherein the difference value is positively correlated with the degree of difference; The step of selecting K recommended scenes from the target scenes corresponding to the difference evaluation indexes according to the total number of recommended scenes and the target scene duty ratio corresponding to the difference evaluation indexes to obtain a recommended scene set including K recommended scenes includes: acquiring a preset recommendation upper limit value and a preset recommendation lower limit value corresponding to each difference evaluation index; For each difference evaluation index, sorting the difference values of all target scenes corresponding to the difference evaluation index according to a sorting mode from large to small, taking the target scenes corresponding to the 1 st to N th difference values of the sorting order as all recommended scenes, and adding all the recommended scenes into the recommended scene set, wherein N is a recommended lower limit value corresponding to the difference evaluation index; acquiring the determined scene number of the recommended scene set; if the number of the determined scenes is smaller than the total number of the recommended scenes, determining an actual upper limit value corresponding to each difference evaluation index according to the total number of the recommended scenes, the target scene duty ratio corresponding to each difference evaluation index and the recommended upper limit value corresponding to each difference evaluation index; and aiming at each difference evaluation index, taking the target scenes corresponding to the (N+1) -th to M-th difference values in the sorting order of the difference values of each target scene corresponding to the difference evaluation index as each recommended scene, and adding each recommended scene into the recommended scene set, wherein M is the actual upper limit value corresponding to the difference evaluation index.
3. The simulation test scene recommendation method according to claim 1, wherein each of the difference evaluation indexes includes a shift difference index; The step of determining the difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index comprises the following steps: According to the gear information in the first automatic driving data, first gear character string information corresponding to each simulation test scene is generated respectively; generating second gear character string information corresponding to each simulation test scene according to the gear information in the second automatic driving data; and calculating a first editing distance between the first gear character string information corresponding to the simulation test scene and the second gear character string information corresponding to the simulation test scene aiming at each simulation test scene, and taking the first editing distance as a difference value of the simulation test scene under the gear difference index.
4. The simulation test scene recommendation method according to claim 1, wherein each of the difference evaluation indexes includes a turn signal difference index; The step of determining the difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index comprises the following steps: respectively generating first turn light character string information corresponding to each simulation test scene according to the turn light state in the first automatic driving data; Respectively generating second turn light character string information corresponding to each simulation test scene according to the turn light state in the second automatic driving data; And aiming at each simulation test scene, calculating a second editing distance between the first steering lamp character string information corresponding to the simulation test scene and the second steering lamp character string information corresponding to the simulation test scene, and taking the second editing distance as a difference value of the simulation test scene under the steering lamp difference index.
5. The simulation test scene recommendation method according to any one of claims 1 to 4, wherein each of the difference evaluation indexes includes an acceleration difference index and a lateral movement difference index; The step of determining the difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index comprises the following steps: And aiming at each simulation test scene, adopting a sliding window method to process the acceleration of the first automatic driving data in the simulation test scene and the acceleration of the second automatic driving data in the simulation test scene to obtain each acceleration difference average value, taking the maximum acceleration difference average value as a difference value of the simulation test scene under the acceleration difference index, and adopting the sliding window method to process the steering wheel rotation angle of the first automatic driving data in the simulation test scene and the steering wheel rotation angle of the second automatic driving data in the simulation test scene to obtain each rotation angle difference average value, and taking the maximum rotation angle difference average value as a difference value of the simulation test scene under the sideslip difference index.
6. The simulation test scene recommendation method according to any one of claims 1 to 4, wherein each of the difference evaluation indexes includes a travel track difference index; The step of determining the difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index comprises the following steps: And for each simulation test scene, calculating the Euclidean distance according to the driving position corresponding to the simulation test scene in the first automatic driving data and the driving position corresponding to the simulation test scene in the second automatic driving data, and taking the Euclidean distance as the difference value of the simulation test scene under the driving track difference index.
7. A simulation test scenario recommendation apparatus, the apparatus comprising: The automatic driving data acquisition module is used for acquiring first automatic driving data output by a first automatic driving algorithm under each simulation test scene and second automatic driving data output by a second automatic driving algorithm under each simulation test scene; The difference determining module is used for determining a difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index; The target scene determining module is used for screening out simulation test scenes with large difference degrees from the simulation test scenes according to the difference values corresponding to the difference evaluation indexes as target scenes corresponding to the difference evaluation indexes; The recommendation scene determining module is used for determining a recommendation scene set according to target scenes corresponding to the difference evaluation indexes, wherein the recommendation scene set comprises at least one target scene, the recommendation scene determining module specifically comprises the steps of accumulating the number of the target scenes corresponding to the difference evaluation indexes to obtain the total number of the target scenes, determining the target scene ratio corresponding to each difference evaluation index according to the total number of the target scenes and the number of the target scenes corresponding to each difference evaluation index, obtaining the preset total number of the recommendation scenes, and selecting K recommendation scenes from the target scenes corresponding to the difference evaluation indexes according to the total number of the recommendation scenes and the target scene ratio corresponding to the difference evaluation indexes to obtain the recommendation scene set comprising K recommendation scenes, wherein K is the total number of the recommendation scenes.
8. A storage medium having stored therein computer readable instructions which, when executed by one or more processors, cause the one or more processors to perform the steps of the simulation test scenario recommendation method of any one of claims 1 to 6.
9. A computer device includes one or more processors and a memory; Stored in the memory are computer readable instructions which, when executed by the one or more processors, perform the steps of the simulation test scenario recommendation method of any one of claims 1 to 6.

Description

Simulation test scene recommendation method and device, storage medium and computer equipment Technical Field The present application relates to the field of autopilot technology, and in particular, to a method and apparatus for recommending a simulation test scenario, a storage medium, and a computer device. Background With continued iteration of autopilot algorithms, autopilot algorithms develop more and more versions. To evaluate the merits of different versions of autopilot algorithms, AB testing of two different versions of autopilot algorithms is required. In the AB test process, two versions of automatic driving algorithms need to be subjected to simulation test under the same simulation test environment and simulation test scene set, and the performance advantages and disadvantages of the two versions are compared according to the automatic driving data output by the two versions of automatic driving algorithms under different simulation test scenes. However, when performing the AB test, a large number of simulation test scenes are often used, but in the prior art, the performance of the automatic driving algorithm with different versions under each simulation test scene is verified one by adopting a manual verification mode, so that the consumption time is long and the test efficiency is low. Disclosure of Invention The application aims to at least solve one of the technical defects, especially the technical defects of long time consumption and low test efficiency in the prior art. In a first aspect, an embodiment of the present application provides a simulation test scenario recommendation method, where the method includes: acquiring first automatic driving data output by a first automatic driving algorithm under each simulation test scene and second automatic driving data output by a second automatic driving algorithm under each simulation test scene; Determining a difference value of each simulation test scene under each difference evaluation index according to the first automatic driving data, the second automatic driving data and each preset difference evaluation index; Aiming at each difference evaluation index, screening out a simulation test scene with large difference degree from the simulation test scenes according to each difference value corresponding to the difference evaluation index as a target scene corresponding to the difference evaluation index; and determining a recommended scene set according to the target scenes corresponding to the difference evaluation indexes, wherein the recommended scene set comprises at least one target scene. In one embodiment, the step of determining the recommended scene set according to the target scenes corresponding to the respective difference evaluation indexes includes: Accumulating the number of the target scenes corresponding to the difference evaluation indexes to obtain the total number of the target scenes; respectively determining the target scene duty ratio corresponding to each difference evaluation index according to the total number of the target scenes and the number of the target scenes corresponding to each difference evaluation index; acquiring the total number of preset recommended scenes; selecting K recommended scenes from the target scenes corresponding to the difference evaluation indexes according to the total number of the recommended scenes and the target scene occupation ratio corresponding to the difference evaluation indexes to obtain a recommended scene set comprising K recommended scenes, wherein K is the total number of the recommended scenes. In one embodiment, the difference value is positively correlated with the degree of difference; The step of selecting K recommended scenes from the target scenes corresponding to the difference evaluation indexes according to the total number of recommended scenes and the target scene duty ratio corresponding to the difference evaluation indexes to obtain a recommended scene set including K recommended scenes includes: acquiring a preset recommendation upper limit value and a preset recommendation lower limit value corresponding to each difference evaluation index; For each difference evaluation index, sorting the difference values of all target scenes corresponding to the difference evaluation index according to a sorting mode from large to small, taking the target scenes corresponding to the 1 st to N th difference values of the sorting order as all recommended scenes, and adding all the recommended scenes into the recommended scene set, wherein N is a recommended lower limit value corresponding to the difference evaluation index; acquiring the determined scene number of the recommended scene set; if the number of the determined scenes is smaller than the total number of the recommended scenes, determining an actual upper limit value corresponding to each difference evaluation index according to the total number of the recommended scenes, the target scene duty ratio correspond