The 11th seminar of the CSIE Seminar Series focused on ensuring safer and more efficient operation of autonomous vehicles in the event of AI system errors.

The seminar was delivered by CSIE Junior Researcher Artyom Khachatryan.

Autonomous vehicles rely heavily on AI for detecting obstacles and making control decisions, but AI systems can occasionally make unpredictable errors. Existing safety systems have addressed this by simply disabling the AI in risky situations, often resulting in unnecessary, abrupt braking.

The talk presented the Synergistic Simplex approach, in which the AI and the safety system operate together: the AI shares important information with the safety system, which in turn guides the AI's decisions, all while guaranteeing the vehicle's safe operation.

Experiments showed that this cooperative approach is able to prevent potential collisions, eliminate unnecessary braking caused by false alarms, and ensure more efficient operation of the autonomous system.