Enhanced parameter estimation in adaptive control via online historical data

Abstract

Parameter convergence is desirable for adaptive control as it enhances the overall stability and robustness properties of the closed-loop system. In existing online historical data (OHD)-driven parameter estimation schemes, all OHD are exploited to update parameter estimates such that exponential parameter convergence is ensured under a condition of sufficient excitation which is strictly weaker than the traditional persistent excitation (PE) condition. However, the exploitation of all OHD not only results in possible unbounded adaptation but also loses the flexibility of handling slow time-varying uncertainties. In this brief, a novel OHD-driven parameter estimation scheme that exploits only partial OHD is presented to improve parameter convergence and is incorporated with direct adaptive control to construct a composite learning control strategy. The proposed approach guarantees exponential parameter convergence under a condition of interval excitation which is also strictly weaker than the PE condition while eliminating the drawbacks of existing OHD-driven parameter estimation schemes. Numerical results have verified the effectiveness and superiority of the proposed approach.