|Verification Futures 2022 (click here to see full programme)
|Agency-directed test generation of AV testing in simulation
|Simulation-based verification is beneficial for assessing dangerous or costly on-road testing of autonomous vehicles (AV). This talk addresses the challenge of efficiently generating effective tests for simulation-based AV verification using software testing agents. I will discuss the properties of a good test case and how our automated method can be used to bias conditions within a simulation towards finding effective tests, edge cases or even test that require sequences. Our results suggest that generating tests using agency-directed testing significantly improves upon random and simultaneously provides more realistic driving scenarios.
|Greg Chance gained a PhD in physics from the University of Bath 2005. He was previously at the Bristol Robotics Lab and is currently a Senior Research Associate with the Trustworthy Systems Laboratory, University of Bristol. His current research interests are simulation-based verification, exploring trustworthiness in autonomous systems and how they can be verified. He is currently working on the EPSRC TAS Functionality node project with the Bristol Robotics Lab working with swarm robots, soft robots and UAVs. He has ten years industrial experience researching for Oxford Instruments, and BAE Systems. Dr Chance is a Chartered Engineer and a member of the Institute of Physics.
- Using agency for test generation is novel
- Agency technique significantly faster at finding tests than random
- Effective use of AI to automate test generation