Multi-agent reinforcement learning: independent vs. cooperative agents
Readings in agents
Cooperative Mobile Robotics: Antecedents and Directions
Autonomous Robots
Coordination in multiagent reinforcement learning: a Bayesian approach
AAMAS '03 Proceedings of the second international joint conference on Autonomous agents and multiagent systems
Predicting and preventing coordination problems in cooperative Q-learning systems
IJCAI'07 Proceedings of the 20th international joint conference on Artifical intelligence
Hi-index | 0.00 |
In distributed virtual environment, through learning, individual CGA(Computer Generated Actor) can adapt environment and other CGA in team, so the team capability of solving problems, the adaptability and robust of CGA team have been increased. When the learning based on random games of team CGA has multiple equilibriums, the equilibrium selection problem of every member in team must be solved. This paper gives a learning method for team CGA called TCCLA. It divides the learning into two levels: managerial member learning and non-managerial member learning. Every member in team selects its optimization actions according to its preference. Non-managerial member learns the optimization equilibrium under the direction of managerial member, so the problem of equilibrium selection has been solved. The IPL algorithm has been improved. The high efficiency of TCCLA has been verified through experiment.