Adaptive Neural Control For Rigid Robotic Manipulators With State Constraints

The robotic manipulator is one of the most commonly used robots in the industry.Depending on whether the end effector is in contact with the operating object,the working of the robotic manipulators can be divided into two types,constrained and unconstrained.The work of a constrained robot is limited by the environment,so it is necessary to restrict its position,speed,and other boundaries.The controller design of state-constrained robotic manipulators is a challenging and hot topic in the field of robot control.This thesis focuses on adaptive neural control for rigid robotic manipulators with full state constraints.The controller design scheme is comprehensively developed for four cases of the constrained robotic manipulator,namely,whether the model is accurately known and whether the external disturbance is considered.Firstly,according to the actual constraints of the position and speed of the manipulator,a new state transition function is proposed to subtly convert the constrained robotic manipulator into a new unconstrained robotic manipulator.And then,the control design is carried out for the converted robotic manipulator system.The specific contents are summarized as follows:For the robotic manipulator with the precisely known model,the second chapter proposes a novel control scheme for the full-state-constraints robotic manipulator by combining the backstepping with the dynamic surface control.The proposed control scheme not only avoids the complexity explosion caused by the traditional backstepping technique due to repeatedly derivation of virtual controllers,but also ensures the trajectory tracking of the end position of the robotic manipulator in the safe working range.In the third chapter,under the condition that the manipulator model is unknown,radial basis function(RBF)neural networks are used to accurately model the partial unknown dynamics of the system.In order to solve the difficulty caused by the unknown controller gain on the stability analysis,we employ the inherent structural characteristics of the robotic manipulator system,and finally achieve the control target successfully.The obtained control effect is similar to the accurately known case of the robotic manipulator model.In the fourth chapter,for the case where the robotic manipulator system model is precisely known but contains unknown external disturbances,a disturbance observer is designed in the forward channel of each joint control input,and combined with dynamic surface control technology,a novel control scheme is proposed for the robotic manipulator with full state constraints and external disturbances.The proposed control scheme not only effectively compensates the impact of disturbances on the control performance,but also achieves the high-precision tracking control of the robotic manipulators system with full state constraints.In the fifth chapter,considering the existence of structural uncertainties and external disturbances,the adaptive neural network control scheme based on disturbance observer is proposed for the full-state-constraints robotic manipulators.In order to solve the difficulty of designing the disturbance observer resulting from the unknown inertia matrix of the robotic manipulators system,we innovatively propose an augment system dynamics technique based on the first-order filter design.By combining dynamic surface control and neural networks,a novel adaptive neural control scheme is proposed for the robotic manipulators subject to full state constraints and external disturbances.The scheme achieves the high-precision tracking control of the robotic manipulators system,meanwhile the full state constraints are not violated.The effectiveness of the proposed schemes is verified by theoretical analysis and MATLAB simulation in each chapter.