Composite adaptive neural tracking control of uncertain strict-feedback systems

Composite adaptive neural control of uncertain strict-feedback systems is synthesized. It mainly consists of an adaptive neural backstepping controller, an immersion and invariance (I&I) update algorithm, and a set of state filters for conquering the in-feasibility of the states' derivatives required in the update algorithm. For compensating the indefinite coupling terms in the update algorithm, a novel nonlinear sigma-modification method is promoted. To further tackle the case with unknown input gain functions, the dynamic surface control (DSC) and the adding-an-integrator technique are incorporated for preventing the so-called explosion of complexity and the algebraic-loop problems in the virtual controller and the composite update algorithm, respectively. The proposed design ensures the semi-globally uniformly ultimately bounded (SGUUB) stability of the closed-loop system and, in particular, the convergence of prediction errors to the vicinity of zero without persistent excitation (PE), which in turn improves the tracking performance.