Algorithms of adaptive disturbance compensation in linear systems with arbitrary input delay

Abstract

ABSTRACT The problem of direct adaptive compensation of a priori uncertain disturbances for general class of linear time-invariant plans with known parameters and arbitrarily known input delay is considered. The external disturbance is assumed to be unmeasurable and modelled as the output of uncertain autonomous linear system. A scheme with two different algorithms of adaptive disturbance compensation is proposed. The scheme is based on special disturbance observer and predictor. The first algorithm involves the conventional gradient algorithm of adaptation with restricted convergence properties. The second algorithm represents the modification of gradient algorithm and accumulates information about disturbance over past period of time. Accumulation is implemented via a preselected linear operator with memory and gives dramatically improved parametric convergence. For algorithms of adaptation, the influence of input delay is completely compensated by the use of a special scheme of error augmentation. The problem of plant stabilisation is resolved by involving the state predictor. The performance of the proposed schemes of disturbance compensation is demonstrated via simulation and compared in conclusion.