CAJ | 학술논문

针对一类非线性系统，提出了具有初态学习的开闭环PD型迭代学习算法，并给出了该算法的收敛充分条件。依据此收敛条件，可确定初态学习律和输入学习律的学习增益，而不必依赖系统的结构和参数，从而放宽了对初始定位的要求。初态学习允许在每次迭代开始时，其初态与期望初态有一定的定位误差，并允许初态在收敛条件范围内任意设置。利用压缩映射分析方法，证明了系统在任意初态下经过几次迭代后，实际输出能完全跟踪上期望轨迹。最后，通过仿真实例验证了所提算法的有效性和可行性。
침대일류비선성계통，제출료구유초태학습적개폐배PD형질대학습산법，병급출료해산법적수렴충분조건。의거차수렴조건，가학정초태학습률화수입학습률적학습증익，이불필의뢰계통적결구화삼수，종이방관료대초시정위적요구。초태학습윤허재매차질대개시시，기초태여기망초태유일정적정위오차，병윤허초태재수렴조건범위내임의설치。이용압축영사분석방법，증명료계통재임의초태하경과궤차질대후，실제수출능완전근종상기망궤적。최후，통과방진실례험증료소제산법적유효성화가행성。
In this paper ,for a class of nonlinear system ,an open‐closed‐loop PD‐type iterative learn‐ing principle with initial state learning is proposed and the sufficient condition for convergence is put for‐warded .Based on this convergence condition ,without depending on the system structure and parame‐ters ,the learning gains of initial state learning law and input learning law can be determined to relax the requirement on initial position .The initial state learning principle allow s a certain degree of orientation bias between actual initial state and desired initial state at the beginning of iteration ,and the actual ini‐tial state can be set arbitrarily in the convergence condition .Using the contraction mapping method ,it is proved that the output of the system with an arbitrary initial state can track the expected trajectory com‐pletely after several iterations .Finally ,the simulation results testify the proposed algorithm is effective and feasible .