Research On Intelligent Control Of Nonlinear Systems Subject To Nonsmooth Actuator Constraints

In automatic control systems, actuator component is generally used to execute control command signal on the plant to achieve the control objective of output regulation or trajectory tracking. In most cases, such a component easily suffers from some nonsmooth constraints, such as the dead-zone effect in transistor, the hysteresis phenomenon in piezoelectrical actuation platform, the backlash constraint in gear transmission system, etc. Due to the existence of these nonlinear constraints, the actuator output is no longer equal to its input, which implies that the tracking performance and stability of the closed-loop system may be destroyed severely if these constraints are not handled well in controller design. Nevertheless, how to compensate for these actuator constraints is a nontrivial task as they are usually nonsmooth and nonlinear to controller design. Thus,it is of both theoretical and practical importance to address such a challenging issue in the field of adaptive control.In this dissertation, the Lyapunov functional approach is used for stability analysis, and the backstepping method is adopted for controller design. By further combining with the techniques of adaptive control, robust control, fuzzy logic systems, neural networks, etc, we successfully solve the following nonlinear control problems.(1) An adaptive neural control scheme is proposed for a class of time-delayed uncertain nonlinear system subject to unknown direction actuator hysteresis. First, a Bouc-Wen model is introduced to describe nonsmooth actuator hysteresis. Next we decompose the model into the form of a control part (control signal with unknown control gain) plus a bounded disturbance-like term such that the well-known Nussbaum approach can be utilized in control design to address the unknown control gain problem. By further constructing a new type of Lyapunov-Krasovskii functional to compensate for time delay effect, the closed-loop system stability in the sense of signal boundedness is successfully established.(2) Two intelligent control schemes are proposed for the output tracking control problem of uncertain nonlinear systems with asymmetric actuator backlash. The first is named as adaptive inverse compensation scheme as it is developed on the newly constructed backlash inverse model and the well-known adaptive approach. Note that such a scheme has two major restrictions, that is, the dependence on known slope ratio and the possible singular problem caused by some adaptive parameters. Therefore, a fuzzy adaptive inverse compensation method is further proposed to remove these restrictions.(3) A robust adaptive neural control scheme is proposed for a class of uncertain nonlinear systems with actuator backlash and quantized control input. We firstly let the real control signal pass the designed hysteresis quantizer before it is transmitted to actuator components via networks, in order to reduce the communication rate of control signal.In this case, the controller design becomes rather challenging as the discrete signal and contionuous signal coexist in the closed-loop control system. To address this issue, a novel decomposition of hysteresis quantizer is proposed by using the sector-bounded property of quantization function. Furthermore, instead of constructing backlash inverse model, we raise a new and novel way which helps to decompose asymmetric backlash function into a proper form for controller design. Based on such decompositions, the real controller is successfully separated from the dynamics of backlash constraint and quantization. By combining with currently available robust design methods, the desired adaptive quantized control scheme is finally developed.(4) An adaptive compensation control scheme is proposed for a class of parametric strict feedback nonlinear systems whose actuator suffers from both nonsmooth constraint and failures/faults. We firstly construct an adaptive inverse compensator to compensate for the damaging effect caused by backlash constraint, and then design a proper input signal of backlash compensator to accommodate the failure effect. With the operations,the desired controller, i.e., the compensator output signal, is obtained. It is rigorously proved via mathematical analysis that the all the closed-loop signal are bounded and the tracking error is convergent to zero.