What is it about?
Autonomous Driving Systems (ADSs) are safety-critical systems, and safety violations of Autonomous Vehicles (AVs) in real traffic will cause huge losses.We introduce motif pattern based on atomic driving maneuvers, which can challenge ADSs effectively. In order to test the performance of ADS comprehensively during long-mile driving, we design a spatiotemporally continuous simulation testing technique, which can continuously create perturbations to the AV.
Featured Image
Photo by Remy Gieling on Unsplash
Why is it important?
We formulate our ADS testing approach as a multi-objective search technique, considering the degree of perturbations to AV and the diversity of scenarios. The motif pattern increases the chance to disturb the ADS during driving. We design a spatiotemporally continuous driving environment, which can test the ADS comprehensively during the long-mile driving
Perspectives
This article explains a testing framework of simulation test for Autonomous Driving Systems. The proposed technique can generate adversarial and diverse testing scenarios by motif pattern and multi-objective evolutionary search, and can test ADSs continously in long-mile driving. The motif pattern can be extended and the multi-objective fitness function can be improved for further research on ADS testing.
Haoxiang Tian
University of the Chinese Academy of Sciences
Read the Original
This page is a summary of: MOSAT: finding safety violations of autonomous driving systems using multi-objective genetic algorithm, November 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3540250.3549100.
You can read the full text:
Contributors
The following have contributed to this page
