Episode 8 (English) - Safety-Validation of autonomous systems (using Machine Learning) - with Anthony Corso

Anthony Corso is a Ph.D. student in the Aeronautics and Astronautics Department at Stanford University where he is advised by Professor Mykel Kochenderfer in the Stanford Intelligent Systems Laboratory (SISL). He studies approaches for the validation of safety-critical autonomous systems with an emphasis on interpretability and scalability. In this podcast he talked about safety-validation of autonomous systems. The latter includes systems such as robots, cars, aircraft, and planetary rovers equally. In May he published a paper which deals with different algorithms for black-box safety validation. One of the approaches is to use reinforcement learning, which was discussed in the podcast in more detail. He also briefly introduced the Next-Generation Airborne Collision Avoidance System ACAS X, in which development Professor Kochenderfer was heavily involved. ACAS X takes advantage of Dynamic Programming, an algorithm for optimal decision making. The mentioned papers, further readings and an interesting podcast can be found here: The paper mentioned above: A Survey of Algorithms for Black-Box Safety Validation The paper on Adaptive Stress Testing (AST): Adaptive stress testing with reward augmentation for autonomous vehicle validation The AST toolbox mentioned in the podcast: AST-Toolbox The CARLA simulator mentioned in the podcast: CARLA A paper on ACAS X: Next-Generation Airborne Collision Avoidance System An episode of Standford University's podcast The Future of Everything with Mykel Kochenderfer where he talkes about ACAS X and Artificial Intelligence (AI) in safety-critical systems: Mykel Kochenderfer: AI and safety-critical systems