Distributionally Robust Inverse Reinforcement Learning for Identifying Multi-Agent Coordinated Sensing
This work addresses the challenge of identifying coordinated behaviors in multi-agent systems, such as radar networks, but is incremental as it builds on existing IRL methods with robustness enhancements.
The paper tackles the problem of reconstructing utility functions in multi-agent sensing systems from noisy observations by developing a distributionally robust inverse reinforcement learning algorithm, achieving robust estimation through a minimax approach and demonstrating efficacy in numerical studies of a cognitive radar network.
We derive a minimax distributionally robust inverse reinforcement learning (IRL) algorithm to reconstruct the utility functions of a multi-agent sensing system. Specifically, we construct utility estimators which minimize the worst-case prediction error over a Wasserstein ambiguity set centered at noisy signal observations. We prove the equivalence between this robust estimation and a semi-infinite optimization reformulation, and we propose a consistent algorithm to compute solutions. We illustrate the efficacy of this robust IRL scheme in numerical studies to reconstruct the utility functions of a cognitive radar network from observed tracking signals.