基于强化学习的多移动Agent学习算法  

Multi Mobile Agent Learning Algorithm Based on Reinforcement Learning

作　　者：刘菲[1] 曾广周[1] 

机构地区：[1]山东大学计算机科学与技术学院,济南250061

出　　处：《计算机工程与应用》2006年第5期50-53,共4页Computer Engineering and Applications

基　　金：国家自然科学基金资助项目(编号:60473123)

摘　　要：结合强化学习技术讨论了单移动Agent学习的过程,然后扩展到多移动Agent学习领域,提出一个多移动Agent学习算法MMAL(MultiMobileAgentLearning)。算法充分考虑了移动Agent学习的特点,使得移动Agent能够在不确定和有冲突目标的上下文中进行决策,解决在学习过程中Agent对移动时机的选择,并且能够大大降低计算代价。目的是使Agent能在随机动态的环境中进行自主、协作的学习。最后,通过仿真试验表明这种学习算法是一种高效、快速的学习方法。Single mobile agent learning progress is discussed based on reinforcement learning technology.And then it is extended to multi mobile agent learning.At last a multi mobile agent learning arithmetic（MMAL） is proposed.MMAL fully takes the characteristic of mobile agent learning into account.It can make agent make a decision in uncertain context with conflict target.Also it can solve the problems that how can an agent choose moving occasion and it can greatly reduce the computational cost.The motive of the algorithm is that mobile agent can study in stochastic dynamic environments independently and concurrently.Simulate results indicate that the proposed processing based on reinforcement learning and clustering is an efficient and fast method.

关 键 词：强化学习 移动Agent学习 学习算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]