A Learning PID Control using Learning Automata

カテゴリ: 部門大会

論文No: MC4-3

グループ名: 【C】平成18年電気学会電子・情報・システム部門大会講演論文集

発行日: 2006/09/05

タイトル(英語): A Learning PID Control using Learning Automata

著者名: 沈方暉 (早稲田大学),庄司文啓 (福岡工業大学),李羲頡 (早稲田大学)

著者名(英語): Fanghui Shen(Waseda University),Fumihiro Shoji(Fukuoka Institute of Technology),Hee-Hyol Lee(Waseda University)

キーワード: Stochastic automaton|Learning|PPID Control|Self-modification

要約(日本語): This paper proposes an implementation of a stochastic automaton for the PID control system with embedded learning function. This implementation would enable PID parameters to be self-modified whenever the characteristic values of controlled objects are changed by disturbance or a change over the years. These self-modified parameters are a system gain, a time constant and a dead time etc. The self-modification mechanism is implemented by applying the self-learning stochastic automata to each of the PID parameters, such as the proportional gain K_p, the integral time T_i, and the derivative time T_d. In order to avoid receiving penalty from the controlled object, unknown dynamical environment, the self-learning stochastic automata can modify to update state transition probability by using a reinforced algorithm, and select outputs based on penalty for the probability of receiving to be small. Simulations have showed that the proposed method is very effective.

Keywords: Stochastic automaton, Learning, PID control, Self-modification

PDFファイルサイズ: 3,092 Kバイト