现代电子技术
現代電子技術
현대전자기술
MODERN ELECTRONICS TECHNIQUE
2014年
4期
1-4
,共4页
PID%参数整定%强化学习%控制系统
PID%參數整定%彊化學習%控製繫統
PID%삼수정정%강화학습%공제계통
PID%parameter tuning%reinforcement learning%control system
控制系统的响应特性取决于控制律参数,经典的 PID 方法难以实现参数的自整定。强化学习能够通过系统自身和环境的交互实现参数的自动调整,但是在控制律参数需要频繁调整的应用场合,常规的强化学习方法无法满足实时性要求,而且容易陷入局部收敛。对传统的强化学习方法加以改进后,加快了在线学习速度,提高了强化学习算法的寻优能力。仿真结果表明,该方法可以在一定范围内快速求得全局最优解,提高控制系统的自适应性,为控制系统参数的自整定提供了依据。
控製繫統的響應特性取決于控製律參數,經典的 PID 方法難以實現參數的自整定。彊化學習能夠通過繫統自身和環境的交互實現參數的自動調整,但是在控製律參數需要頻繁調整的應用場閤,常規的彊化學習方法無法滿足實時性要求,而且容易陷入跼部收斂。對傳統的彊化學習方法加以改進後,加快瞭在線學習速度,提高瞭彊化學習算法的尋優能力。倣真結果錶明,該方法可以在一定範圍內快速求得全跼最優解,提高控製繫統的自適應性,為控製繫統參數的自整定提供瞭依據。
공제계통적향응특성취결우공제률삼수,경전적 PID 방법난이실현삼수적자정정。강화학습능구통과계통자신화배경적교호실현삼수적자동조정,단시재공제률삼수수요빈번조정적응용장합,상규적강화학습방법무법만족실시성요구,이차용역함입국부수렴。대전통적강화학습방법가이개진후,가쾌료재선학습속도,제고료강화학습산법적심우능력。방진결과표명,해방법가이재일정범위내쾌속구득전국최우해,제고공제계통적자괄응성,위공제계통삼수적자정정제공료의거。
The response characteristics of control system depend on the control law parameter.The classic PID method is dif-ficult to achieve the parameter self-tuning.Through the interaction of system itself and the environment,parameters can be adjusted automatically by reinforcement learning.However,in the application occasions where the control law parameters requires to be ad-justed frequently,the conventional reinforcement learning methods cannot meet the real-time requirements,and is easy to fall in-to local convergence.Based on the traditional reinforcement learning methods,an improvement method which can accelerate the learning speed and improve the optimizing ability of reinforcement learning algorithm is proposed.The simulation results show that this method can get global optimal solution quickly and improve the adaptivity of the control system in a certain range.It pro-vided a basis for the improvement of control system’s parameter self-tuning.