To Switch or Not to Switch? Balanced Policy Switching in Offline Reinforcement Learning
This addresses a key challenge in offline RL for decision-making systems like robotics and traffic control, offering a principled approach to policy switching, though it is incremental in applying optimal transport ideas to this specific scenario.
The paper tackles the problem of balancing the gain and cost of policy switching in offline reinforcement learning, where switching incurs a cost and only historical data is available. It introduces a novel formulation using optimal transport and demonstrates efficiency on robot control and traffic light benchmarks.
Reinforcement learning (RL) -- finding the optimal behaviour (also referred to as policy) maximizing the collected long-term cumulative reward -- is among the most influential approaches in machine learning with a large number of successful applications. In several decision problems, however, one faces the possibility of policy switching -- changing from the current policy to a new one -- which incurs a non-negligible cost, and in the decision one is limited to using historical data without the availability for further online interaction. Despite the inevitable importance of this offline learning scenario, to our best knowledge, very little effort has been made to tackle the key problem of balancing between the gain and the cost of switching in a flexible and principled way. Leveraging ideas from the area of optimal transport, we initialize the systematic study of policy switching in offline RL. We establish fundamental properties and design a Net Actor-Critic algorithm for the proposed novel switching formulation. Numerical experiments demonstrate the efficiency of our approach on multiple robot control benchmarks of the Gymnasium and traffic light control from SUMO-RL.