[WAFR'24] Stackelberg Adversarial RL for Convergent Neural Synthesis of Robot Safety

The 3rd MAD-Games workshop at ACC 2026 aims to explore the latest advances in using game-theoretic and multi-agent control and learning approaches to help autonomous agents achieve safe interactions in highly dynamic environments. It brings together researchers and practitioners to (1) explore the latest developments in interactive autonomy, e.g., game theory, distributed control, and multi-agent learning, and (2) investigate how game-theoretic models can be adapted to handle the complexities of real-world systems.

In this course, we will study fundamental concepts in human-centered robotics, with an emphasis on mathematical models of human-robot interaction and decision-making algorithms for safely deploying robots in human-populated environments.

We consider a coaching setting where an expert AI teaches a novice human to acquire motor skills for high-speed drone racing. We formulate this problem as a general-sum game, where the AI's goal is to promote long-term human skill development and independence. By strategically reducing AI assistance and providing demonstrations after the human has experienced failures, the AI coach can foster more effective learning than with static, full assistance. In this case, the AI is no longer a passive guardian angel but an active mentor that deliberately orchestrates challenging scenarios as structured learning signals.