Low-Frequency Learning and Fast Adaptation in Least Squares Adaptive Control with Normalized Variable-Damping Coefficient

This paper proposes a novel least squares adaptive control architecture to improve the performance against nonlinear systems with mismatched uncertainties. To address the poor transient of the predictive model-based least squares (P-LS) adaptive control, the designed predictive model-based normalized variable-damping least squares filtered (P-VLSF) adaptive control obtained the capability of low-frequency learning and fast adaptation by incorporating a filter-modified term in the adaptive law. The critical feature of the proposed P-VLSF is that it contains a normalized variable-damping coefficient, which helps achieve improved transient performance with a reduced cost of error tracking performance. And the transient and steady-state performance is guaranteed theoretically. Furthermore, the efficacy of the proposed scheme has been demonstrated by flight control simulations. In particular, it is also verified that P-VLSF has better tracking performance than predictive model-based least squares filtered (P-LSF) adaptive control when fast adaptation, which is attributed to the design of the normalized variable-damping coefficient.