Reinforcement learning methods are not yet widely used in computer games, at least not for demanding online learning tasks. This is in part because such methods often require excessive number of training samples before converging. This can be particularly troublesome in mobile game devices where both storage and CPU are limited and valuable resources. In this paper we describe a new AI-based game for mobile phones that we are currently developing. We address some of the main challenges of incorporating efficient on-line reinforcement learning methods into such gaming platforms. Furthermore, we introduce two simple methods for interactively incorporating user feed-back into reinforcement learning. These methods not only have the potential of increasing the entertainment value of games, but they also drastically reduce the number of training episodes needed for the learning to converge. This enhancement made it possible for us to use otherwise standard reinforcement learning as the core part of the learning AI in our game.
[1]
Ben J. A. Kröse,et al.
Learning from delayed rewards
,
1995,
Robotics Auton. Syst..
[2]
Richard S. Sutton,et al.
Reinforcement Learning: An Introduction
,
1998,
IEEE Trans. Neural Networks.
[3]
Gerald Tesauro,et al.
TD-Gammon, a Self-Teaching Backgammon Program, Achieves Master-Level Play
,
1994,
Neural Computation.
[4]
Eric O. Postma,et al.
On-Line Adaptation of Game Opponent AI in Simulation and in Practice
,
2003,
GAME-ON.