摘要
针对Q学习状态空间非常大,导致收敛速度非常慢的问题,给出一种基于边界样本协调的多智能体在线合作学习方法,使得智能体在特定的子空间上进行特化并通过边界状态上的开关函数相互协调,从而能够较快地学习到局部最优.仿真实验表明该方法能够取得比全局学习更好的在线学习性能.
Aiming at the large state-space caused by the slow convergence of Q learning, a kind of multiagent cooperative learning is proposed by the coordination of boundary samples. Each agent is specialized in its sub-space, and the agents coordinate through Boolean functions in boundary states. Simulation results have proved that the proposed method performs better than the traditional global learning.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2008年第1期111-115,共5页
Pattern Recognition and Artificial Intelligence
关键词
多智能体系统
强化学习
多智能体合作
Muhiagent System, Reinforcement Learning, Multiagent Cooperation
