Model-free predictive iterative learning safety-critical consensus control for multi-agent systems with randomly varying trial lengths

Abstract

In this paper, a model-free predictive iterative learning control (MFPILC) is proposed for a class of nonlinear and nonaffine single-input multiple-output (SIMO) discrete-time multi-agent systems with randomly varying trial lengths, input time-varying delay and nonrepetitive external disturbances to deal with the safety-critical consensus control problem. First, the agent dynamics are transformed into a novel dynamic linearization data model along the iteration axis containing the upper and lower bound of the unknown input time delay with an unknown system parameter pseudo gradient (PG). A steepest-descent estimation algorithm updated along the iterative learning axis is applied to deal with the unknown PG and the unknown non-repetitive external disturbances. Further, the predictive compensation mechanisms are proposed to address the problem of data loss caused by the varying trial lengths at each iteration of the MASs. Therefore, a predictive iterative learning consensus control method is designed for MASs based on consensus error and data compensation mechanisms. Moreover, a new model-free predictive iterative learning safety-critical consensus control combined with a discrete-time iterative learning control barrier function (ILCBF) is presented to ensure the output safety of nonlinear MASs. Finally, simulation results further verify the effectiveness of the proposed method.