Abstract
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.
Original language | English (US) |
---|---|
Pages (from-to) | 21-33 |
Number of pages | 13 |
Journal | Proceedings of Machine Learning Research |
Volume | 144 |
State | Published - 2021 |
Event | 3rd Annual Conference on Learning for Dynamics and Control, L4DC 2021 - Virtual, Online, Switzerland Duration: Jun 7 2021 → Jun 8 2021 |
All Science Journal Classification (ASJC) codes
- Artificial Intelligence
- Software
- Control and Systems Engineering
- Statistics and Probability
Keywords
- Causality
- Generalization
- Invariance
- Reinforcement Learning