The Parti-Game Algorithm for Variable Resolution Reinforcement Learning in Multidimensional State-Spaces

Parti-game is a new algorithm for learning feasible trajectories to goal regions in high dimensionalcontinuousstate-spaces. In high dimensions it is essential that learningdoes not plan uniformlyover a state-space. Parti-gamemaintainsa decision-treepartitioningof state-space andappliestechniquesfromgame-theoryandcomputationalgeometryto e cientlyand adaptively concentratehigh resolutiononly on critical areas. The currentversion of the algorithmis designed to nd feasible paths or trajectories to goal regions in high dimensional spaces. Future versions will be designedto nd a solutionthat optimizesa real-valuedcriterion. Many simulatedproblems havebeen tested,rangingfrom two-dimensionalto nine-dimensionalstate-spaces,includingmazes, path planning, non-linear dynamics, and planar snake robots in restricted spaces. In all cases, a good solution is found in less than ten trials and a few minutes.

• Type: Conference
• Year: 1993
• Where: NIPS
• Authors: Andrew W. Moore
• Keywords: Feasible Trajectories, Goal Regions, NIPS 1993, NIPS 2007, Parti-gamemaintainsa Decision-treepartitioningof State-space