Adaptive Neural Network Control For Uncertain Nonlinear Systems With Output Constraints And State Constraints And Its Application

Many practical systems,such as the process industry,power systems and robotic systems,are subject to some kinds of hard constraints,i.e.,some key systems states are required to remain in certain boundary during the transient and steady period.Due to the practical systems are always the nonlinear mathematical model,and the model are usually partial known and even totally unknown,the key states have the possibility to violate these hard constraints by the external disturbances,which result in a series of accidents.Therefore,the research on the constrained control of uncertain nonlinear systems subject to external disturbances is very meaningful to both practical systems and the development of theories.Motivated by the above considerations,this thesis aims at addressing the constrained control design for several typical uncertain nonlinear systems subject to output and state constraints,as well as the external unknown disturbances.To guarantee the constraints satisfaction,this thesis mainly adopts the integral Barrier Lyapunov Functionals(iBLFs)for the control design.Compared with the traditional Lyapunov functions defined in the whole spac and being radially unbounded for the global stability,the Barrier Lyapunov Functions(BLFs)are defined in the constrained subspace,and hold a special property,that is,the function values would grow to infinity when the arguments approach the boundary.Further,the proposed iBLFs in this thesis extend the feasible initial states to the whole constrained space,and are able to handle the unknown control gains.This thesis adopts the proposed iBLFs in the robust adaptive neural network control,and guarantee the boundedness by iBLFs along the trajectory of closed-loop systems by designing appropriate control systems,and the constraints can be guaranteed.The main contributions in the thesis are the follows:Firstly,this thesis considers the single-input single-output strict-feedback nonlinear systems with output constraints and unknown drift system functions.Based on the Backsteppig design,the iBLFs are adopted to deduce the ideal virtual stabilization function in the first step for output constraints,and the remaining steps adopts the quadratic functions.To handle the unknown packaged functions in the virtual stabilization functions,the radial basis function neural networks(RBF NNs)are used to approximate the unknown parts,and adapting parameters are embedded in the control signal to estimate the unknown bounds on the NNs approximations to enhance the robustness.Then to guarantee the satisfaction of full state constraints for strict-feedback nonlinear systems subject to unknown disturbances,new iBLFs are developed to handle both state constraints and unknown control gains,and are brought in each step of Backstepping design.In each design step,the RBF NNs and adapting parameters are also constructed to approximate unknown functions and unknown bounds.In the full state constraints case,the constraints can not be arbitrarily rendered,and they should satisfy the feasibility conditions depend on the initial conditions and design parameters.Accordingly,before the implementation of proposed control,this thesis proposes a feasibility check step to obtain the optimal design parameters to guarantee the feasibility conditions and maximize the tracking performance.To extend the above designs to the non-affine pure-feedback nonlinear systems,under mild general assumptions,this thesis adopts the mean-value theorem to transfer the non-affine system to affine form,and then designs the full state constrained control.Further,to address the state-constrained control under input saturation,this thesis adopts a smooth control signal function to approximate the non-differential saturation function,and obtain the design form by using the mean-value theorem.For the feasibility of full state constrained control subject to the input saturation,a new feasibility condition on the controllability of input saturation systems is added in the feasibility check.Finally,the proposed control design is also utilized in a practical scenario,that is,the trajectory tracking of robotic manipulator under joint space constraints and task space constraints.Based on coupled multi-input multi-output(MIMO)manipulator and actuators models,this thesis utilizes the recursive Backstepping steps to design the control input voltage by using the iBLFs,NNs and constructing the adapting parameters.Based on a two degrees of freedom robotic manipulator with practical parameters,the simulations results show that the manipulator joints as well as the positions of manipulator's end-effector are guaranteed in the predefined constrained space,and are able to practically track the predefined trajectory.