US 12,072,262 B2
Automatic driving acceleration test method considering efficiency and coverage
Bing Zhu, Changchun (CN); Peixing Zhang, Changchun (CN); Jian Zhao, Changchun (CN); Yuhang Sun, Changchun (CN); and Tianxin Fan, Changchun (CN)
Filed by Jilin University, Changchun (CN)
Filed on Sep. 2, 2022, as Appl. No. 17/901,879.
Claims priority of application No. 202111398403.2 (CN), filed on Nov. 24, 2021.
Prior Publication US 2022/0412843 A1, Dec. 29, 2022
Int. Cl. G01M 17/007 (2006.01); B60W 50/04 (2006.01); B60W 60/00 (2020.01)
CPC G01M 17/0078 (2013.01) [B60W 50/045 (2013.01); B60W 60/001 (2020.02); B60W 2520/105 (2013.01)] 7 Claims
OG exemplary drawing
 
1. An automatic driving acceleration test method considering efficiency and coverage, comprising:
Step 1: definition of scenario test priority;
determining scenario hazard, scenario exposure frequency and scenario sensitivity of different specific test scenarios according to natural driving data, and then calculating a test priority wi corresponding to a specific test scenario;
Step 2: zone division;
dividing a scenario generation range parameter space according to the test priority wi of the specific test scenario, and dividing the specific test scenarios with similar test priorities together,
Step 3: search within zones;
selecting specific test scenarios from all the divided zones in turn, forming a set to be tested in this round, and testing a tested autopilot algorithm by the specific test scenarios in the set to be tested to obtain a result;
Step 4: update of scenario test priorities;
comparing actual scenario hazards of the selected scenarios in each zone in the obtained test result with scenario hazards at a location of the specific test scenario parameters obtained by the natural driving data, and updating the test priority of the specific test scenario corresponding to the tested algorithm in the scenario generation parameter space;
Step 5: iterative test;
repeating the steps 2, 3 and 4 until the test priorities of all specific test scenarios remaining in the scenario generation range is lower than a set threshold.