Invariant Policy Optimization: Towards Stronger Generalization in Reinforcement Learning
This work addresses the problem of poor generalization in reinforcement learning for agents operating in diverse environments, offering a method to enhance robustness in applications like robotics.
The paper tackles the challenge of learning reinforcement learning policies that generalize to unseen domains by introducing an invariance principle and a novel algorithm, Invariant Policy Optimization (IPO), which shows significant improvements in generalization performance on tasks like linear quadratic regulator and grid-world problems.
A fundamental challenge in reinforcement learning is to learn policies that generalize beyond the operating domains experienced during training. In this paper, we approach this challenge through the following invariance principle: an agent must find a representation such that there exists an action-predictor built on top of this representation that is simultaneously optimal across all training domains. Intuitively, the resulting invariant policy enhances generalization by finding causes of successful actions. We propose a novel learning algorithm, Invariant Policy Optimization (IPO), that implements this principle and learns an invariant policy during training. We compare our approach with standard policy gradient methods and demonstrate significant improvements in generalization performance on unseen domains for linear quadratic regulator and grid-world problems, and an example where a robot must learn to open doors with varying physical properties.