Model-Based Reinforcement Learning for Control under Time-Varying Dynamics
This addresses the challenge of real-world systems with drift or changing conditions for robotics and control applications, but it is incremental as it builds on existing model-based reinforcement learning methods.
The paper tackles the problem of reinforcement learning for control under time-varying dynamics, which violates the common assumption of stationary systems, and demonstrates improved performance on continuous control benchmarks with non-stationary dynamics.
Learning-based control methods typically assume stationary system dynamics, an assumption often violated in real-world systems due to drift, wear, or changing operating conditions. We study reinforcement learning for control under time-varying dynamics. We consider a continual model-based reinforcement learning setting in which an agent repeatedly learns and controls a dynamical system whose transition dynamics evolve across episodes. We analyze the problem using Gaussian process dynamics models under frequentist variation-budget assumptions. Our analysis shows that persistent non-stationarity requires explicitly limiting the influence of outdated data to maintain calibrated uncertainty and meaningful dynamic regret guarantees. Motivated by these insights, we propose a practical optimistic model-based reinforcement learning algorithm with adaptive data buffer mechanisms and demonstrate improved performance on continuous control benchmarks with non-stationary dynamics.