Robust Reinforcement Learning
- 1 February 2005
- journal article
- Published by MIT Press in Neural Computation
- Vol. 17 (2) , 335-359
Abstract
This letter proposes a new reinforcement learning (RL) paradigm that explicitly takes into account input disturbance as well as modeling errors. The use of environmental models in RL is quite popular for both off-line learning using simulations and for online action planning. However, the difference between the model and the real environment can lead to unpredictable, and often unwanted, results. Based on the theory of H∞ control, we consider a differential game in which a “disturbing” agent tries to make the worst possible disturbance while a “control” agent tries to make the best control input. The problem is formulated as finding a min-max solution of a value function that takes into account the amount of the reward and the norm of the disturbance. We derive online learning algorithms for estimating the value function and for calculating the worst disturbance and the best control in reference to the value function. We tested the paradigm, which we call robust reinforcement learning (RRL), on the control task of an inverted pendulum. In the linear domain, the policy and the value function learned by online algorithms coincided with those derived analytically by the linear H∞ control theory. For a fully nonlinear swing-up task, RRL achieved robust performance with changes in the pendulum weight and friction, while a standard reinforcement learning algorithm could not deal with these changes. We also applied RRL to the cart-pole swing-up task, and a robust swing-up policy was acquired.Keywords
This publication has 8 references indexed in Scilit:
- Robust reinforcement learning control with static and dynamic stabilityInternational Journal of Robust and Nonlinear Control, 2001
- Acquisition of stand-up behavior by a real robot using hierarchical reinforcement learningRobotics and Autonomous Systems, 2001
- Value-function reinforcement learning in Markov gamesCognitive Systems Research, 2001
- Reinforcement Learning in Continuous Time and SpaceNeural Computation, 2000
- Risk-sensitive and minimax control of discrete-time, finite-state Markov decision processesAutomatica, 1999
- Practical issues in temporal difference learningMachine Learning, 1992
- A stochastic reinforcement learning algorithm for learning real-valued functionsNeural Networks, 1990
- Neuronlike adaptive elements that can solve difficult learning control problemsIEEE Transactions on Systems, Man, and Cybernetics, 1983