A quadratic programming formulation of a moving ball interception and shooting behaviour, and its application to neural network control

Abstract

A desirable elementary behaviour for a robot soccer player is the moving ball interception and shooting behaviour, but generating smooth, fast motion for a mobile robot in a changing environment is a difficult problem. We address this problem by formulating the specifications of this behaviour as a quadratic programming optimisation problem, and by training a neural network controller on the exact solution computed offline by a quadratic programming problem optimiser. We present experimental results showing the validity of the approach and discuss potential applications of this approach in the context of reinforcement learning. © 2001 Springer-Verlag Berlin Heidelberg.