Robust higher-order feedback feedforward ILC for networked nonlinear system with varying trial lengths, data dropouts...

This paper investigates iterative learning control (ILC) for networked discrete-time nonlinear systems subject to input-output nonlinearity, iteration-varying trial lengths, random data dropouts, disturbances, and initial state drifts. Unlike many existing ILC results that address only part of these factors, this study develops a robust higher-order feedback feedforward ILC framework that simultaneously handles varying trial lengths and independent random packet dropouts in both the sensor and actuator channels. The calculated and actual inputs are incorporated into the control design to maintain system operation against random data dropouts within the network. To compensate for the missing tracking information induced by iteration-varying trial lengths, a higher-order feedforward learning mechanism is introduced to extract and utilize information from several most recent iterations, thereby improving the use of historical learning data under nonidentical trial durations. Through the method of mathematical induction, it is shown that despite the presence of different trial durations, input-output nonlinearity, random data dropouts and disturbances, the higher-order feedback feedforward ILC method presented in this paper can confine the mathematical expectation of the ILC tracking errors to a bounded region. The extent of this region is related to the disturbances and initial state drifts. More precisely, in the absence of disturbances and initial state drifts, the tracking errors of ILC approach zero when considered from the perspective of mathematical expectation. A simulation example is employed to showcase the efficacy of the proposed algorithm.