Efficient learning of reactive robot behaviors with a Neural-Q/spl I.bar/learning approach
Marc Carreras, Pere Ridao, J. Batlle, Tudor Nicosevici · 2003
The purpose of this paper is to propose a Neural-Q/spl I.bar/learning approach designed for online learning of simple and reactive robot behaviors. In this approach, the Q/spl I.bar/function is generalized by a multi-layer neural network allowing the use of continuous states and actions. The algorithm uses a database of the most recent learning samples to accelerate and guarantee the convergence. Each Neural-Q/spl I.bar/learning function represents an independent, reactive and adaptive behavior which maps sensorial states to robot control actions. A group of these behaviors constitutes a reactive control scheme designed to fulfill simple missions. The paper centers on the description of the Neural-Q/spl I.bar/learning based behaviors showing their performance with an underwater robot in a target following task. Real experiments demonstrate the convergence and stability of the learning system, pointing out its suitability for online robot learning. Advantages and limitations are discussed.