OxBlue2008(3D) Team Description
Jie Ma, Stephen Cameron · 2008
OxBlue2008(3D) is a robot football team for RoboCup 3D simulation. In this paper, the decision structure of our team will be presented, and a walking pattern problem that encountered by most 3D teams will be reviewed. We propose a novel method called PSP (Policy Search Planning), and discuss how PSP can be used to solve the walking pattern problem. We regard joints and parts from a single robot as multi- agents; a walking pattern can be deflned as a cooperative plan. Policy Search is used to flnd an optimal policy for selecting difierent walking patterns from a plan pool; it extends an existing gradient-search method (GPOMDP) to a MAS domain.