A Minimax Approach to Ad Hoc Teamwork
This work addresses robustness in multi-agent systems for scenarios with uncertain teammates, but it is incremental as it builds on existing methods by focusing on worst-case guarantees.
The paper tackles the problem of Ad Hoc Teamwork by proposing a minimax-Bayes approach that optimizes policies against an adversarial prior over partners to improve worst-case performance guarantees, showing superior robustness in experiments on coordinated cooking tasks.
We propose a minimax-Bayes approach to Ad Hoc Teamwork (AHT) that optimizes policies against an adversarial prior over partners, explicitly accounting for uncertainty about partners at time of deployment. Unlike existing methods that assume a specific distribution over partners, our approach improves worst-case performance guarantees. Extensive experiments, including evaluations on coordinated cooking tasks from the Melting Pot suite, show our method's superior robustness compared to self-play, fictitious play, and best response learning. Our work highlights the importance of selecting an appropriate training distribution over teammates to achieve robustness in AHT.