Multi-Agent Interactions Modeling with Correlated Policies
arXiv:2001.03415
Abstract
In multi-agent systems, complex interacting behaviors arise due to the high correlations among agents. However, previous work on modeling multi-agent interactions from demonstrations is primarily constrained by assuming the independence among policies and their reward structures. In this paper, we cast the multi-agent interactions modeling problem into a multi-agent imitation learning framework with explicit modeling of correlated policies by approximating opponents' policies, which can recover agents' policies that can regenerate similar interactions. Consequently, we develop a Decentralized Adversarial Imitation Learning algorithm with Correlated policies (CoDAIL), which allows for decentralized training and execution. Various experiments demonstrate that CoDAIL can better regenerate complex interactions close to the demonstrators and outperforms state-of-the-art multi-agent imitation learning methods. Our code is available at \url{https://github.com/apexrl/CoDAIL}.
20 pages (10 pages of supplementary), 5 figures, Accepted by The Eighth International Conference on Learning Representations (ICLR 2020)
References in corpus (11)
- Counterfactual Multi-Agent Policy Gradients
- Optimizing Neural Networks with Kronecker-factored Approximate Curvature
- Autonomous Agents Modelling Other Agents: A Comprehensive Survey and Open Problems
- Reinforcement Learning with Deep Energy-Based Policies
- Multi-Agent Generative Adversarial Imitation Learning
- Learning Policy Representations in Multiagent Systems
- Multi-Agent Adversarial Inverse Reinforcement Learning
- Coordinated Multi-Agent Imitation Learning
- Multiagent Soft Q-Learning
- Inverse Reinforcement Learning in Swarm Systems
- A Regularized Opponent Model with Maximum Entropy Objective