Please note: This master’s thesis presentation will take place in E7 5419 and online.
Eli-Henry Dykhne, Master’s candidate
David R. Cheriton School of Computer Science
Supervisor: Professor Krzysztof Czarnecki
Dynamic occlusions (occlusions caused by other moving road objects) pose some of the most difficult challenges for autonomous driving systems (ADS). Validating the robustness of ADSs to safety critical dynamic occlusions is a difficult task due to the rarity of such scenarios in recorded driving logs. We provide a systematic typology of dynamic occlusion scenarios involving vehicles, as well as an interpretable framework for ADS safety validation in the presence of dynamic occlusions. Our framework allows for the generation of a diverse set of dynamic occlusion-caused collisions (OCCs) across a wide variety of intersections. These OCCs are tailored directly for the ADS under test. We have implemented our method of generation in a tool (OCC-Gen) and provide results demonstrating that it achieves higher generation efficiency and diversity of OCCs than prior works, while being applicable to any map or intersection. Our tool allows for the custom tailoring of scenario generation, occlusion calculation, and behaviour of non-ego vehicles, as well as driver reaction time. In this work, we present our technique and provide a detailed analysis of the variety and quality of our generated scenarios.
To attend this master’s thesis presentation in person, please go to E7 5419. You can also attend virtually using MS Teams.