Research On Underactuated Manipulator Control System

Nonlinear underactuated control systems are nonlinear systems whose degrees of freedom are less than their independent control variables. Because of the complexity, nonlinearity, instability and multi-objective of underactuated systems, control of nonlinear underactuated systems is very difficult. However, due to its advantages in reducing the number of implementation components, weight and cost, and flexible operation, the study of underactuated systems is significant for control theory and engineering.Underactuated manipulator is a typical nonlinear underactuated mechanical system, which consists of two rigid connecting rods. The first joint is actuated by a DC motor, called actuated link; the second joint connects two rigid rods, called underactuated link. This system has one input variable and two output variables and can be used for classical underactuated experiment research.This thesis investigates underactuated manipulator control system and concerns the control performance, the security of system operation and convenience to operate. This thesis first designs and develops the special control software of underactuated manipulator control system. Then, a LQR balance control algorithm basing on T-S fuzzy control is proposed and neural network is introduced to compensate the friction disturbances, which overcome some drawbacks of the traditional balance control algorithms. Finally, experimental results show the validity and advantage of the proposed control algorithms. The main contributions are summarized are as follows:1. The system structure, characteristics and research function of underactuated manipulator control system are described and the state of the art of the research on underactuated manipulator algorithms is reviewed. The dynamics model of underactuated manipulator is established by the Lagrange dynamics equation and the parameter identification equation is derived by the energy method.2. Based on underactuated manipulator dynamic model, this thesis proposes LQR balance controller basing on T-S fuzzy control and an energy swing up controller basing on neural network compensation. The balance control algorithm consists of decomposing state space into several fuzzy subspaces, designing local LQR controllers and connecting all the local controllers smoothly by fuzzy theory. It is shown that the fuzzy LQR controller can lead to better control performance. In swing up control algorithm, neural network is introduced to compensate the unmodeled dynamics of system to obtain better control performance.3. This thesis presents special control software of underactuated manipulator system, and designs and develops complex algorithm modules, filter module, switch module, virtual animation module and real-time signal detection module, etc, which ensures that the system can run safely, be easy to monitor and control. Furthermore, experimental tests show the validity and advantage of the software. Finally, by the proposed software, the traditional algorithms and the newly proposed algorithm are implemented. The experimental tests show that the newly proposed algorithm leads to better control performance than the traditional algorithms.