Fault-tolerant Control For Reconfigurable And Modular Robots Based On Local Information

With the rapid development of modern industry, the expectancy and requirements for the people to the robots are higher and higher. Owing to the limit of itself machinery structure for conventional robots with fixed configuration, the work scope to be finished is quite limited. Just the working task having been provided, it is to adapt different tasks by adjusting system parameter. However, As to some unknown environment that we can't know the work task beforehand, such as in the space station, lunar base and nucleus scrap heap recycle bin etc. It is very difficult and even unavailable to design such a robot that can finish large scope working requirements. In order to finish these continual changeable working tasks in these unknown environments, it demands to adopt reconfigurable and modular robot that can change its kinematical and dynamical parameters due to the different environment instead of conventional robots. Owing to reconfigurable and modular robot usually being used in the fields of distant, unknown and dangerous, the requirements of its safety and dependence are higher and higher. For example, it demands that reconfigurable and modular robot system has the ability of fault diagnosis and fault-tolerance control. Although scholars at home and abroad have deeply studied the reconfigurable and modular, focus on its configuration optimization, automatic generation in kinematics and dynamics and control method, however, owing to the study of fault diagnosis and fault-tolerance control, just underway. It is very necessary and significant to study on its fault diagnosis and fault-tolerance control.This paper studies on dynamic model of reconfigurable and modular robot, and in connection with actuator failure of every subsystem of reconfigurable and modular robot, fault detection, fault identification, activity fault-tolerance and robust fault-tolerance are proposed regularly. Main content and innovation are as follows.This paper presents an automated method to build dynamic model based on rigid body Newton-Euler iterative algorithm, a iterative Newton-Euler function of the system is got by iterative of generalized velocity generalized acceleration and inverted iterative of generalized force. The dynamics function of reconfigurable and modular robot is set up, and based on the whole dynamic model, it adopts recurrent method to obtain dynamic equation of every subsystem. At last, it gives regular fault model of every subsystem actuator.At first, based on dynamic model of every subsystem of reconfigurable and modular robot, designs RBF neural network adaptive decentralized control, and regards this control method as regular control method when systems are regularly working. Secondly, in connection with actuator failure of every subsystem of reconfigurable and modular robot, a fault detection method based on nonlinear velocity observer and a fault identification method based on state observer. Though the actual value of velocity and observed value of velocity, obtained the residual and compare residual as threshold of system provided. When data is greater than threshold, system actuator occurs faults; on the contrary, systems work regularly, it's up to this aim by making use of this method. And then it regards defined new residual vector and unknown actuator fault function as state variables, by which obtains state equation. When system state is observable, it adopts unknown input observer to estimate unknown actuator fault function of every subsystem so as to achieve the purpose of fault identification. At last, on the basis of fault diagnosis, an active fault tolerant control algorithm for distributed based on RBF neural network compensation is proposed and realize the active fault tolerant control for reconfigurable modular robot of each subsystem in the event of actuator failure. These methods are proved by Lyapunov stability theory to the stability of closed-loop system. Through the simulation results, fault detection method has real-time and strong robustness. Fault identification method has a very good ability to identify actuator faults and verifies when using active fault-tolerant control, even if a fault occurs, this control system can still ensure the stability and tracking accuracy.In connection with parts of the fault of reconfigurable modular robot system actuator, tolerant control method based on inverse design combined with the delay technology is proposed. Owing to this method bringing in the delay control, detection and estimation of the information of actuator failure is not online so as to avoid the possibility of the general fault-tolerant control system when occurs failure diagnosis error and the problem of reconstruction of fault-tolerant control caused by the separation delay of deterioration of control performance is not existed. This control method combines inverse neural network control with delay control technology and it has strong robustness for uncertainties and related items of dynamic model parameters and achieves fault-tolerant control timely when occurs faults. At last, it verifies the effectiveness of its methods through the simulation results.In the end, i summarize the work of the full text and combines with my research experience and i outlook some issues on further research in the future.