Time Scale Risk-Sensitive Hierarchical Structure Control Problem
Hierarchical structure control problem solving is one of the keys to study a controlled Agent who learns its environment efficiently with large state space. It is more and more important to some practical applications in large state space control problems because of requiring an Agent to fit more complex environment, specially, in the area of studying for machine learning. We always regard the Markov Decision Processes (MDP) as the environment model of a reinforcement learning Agent.Bellmans optimal control equation is the base to solving this problem in simulating experiments and practical applications. We have introduced a kind of more complex and more practical environment for learning Agent in our previous work. Combining two concepts of risk-sensitive and multi-time scale, we have proposed a new conception which we refer it to multi-time scale risk-sensitive Markov Decision Processes. Under this new conception,we have modeled the basic Bellmans optimal control equation. Our motivation in this paper is to investigate this problem continually and gives a set of basic results.These results are all cores for framework of solving multi-time scale risk-sensitive control problems.
Hierarchical Structure Control Markov Decision Processes Multi-time Scale Risk Sensitive Bellman Equation
Changming Yin Huanwen Chen Lijuan Xie
College of Computer and Communicational Engineering, Changsha University of Science and Technology C College of Computer and Communicational Engineering, Changsha University of Science and Technology C
国际会议
杭州
英文
990-993
2006-10-12（万方平台首次上网日期，不代表论文的发表时间）