Fault Diagnosis And Tolerant Control For Reconfigurable Robots Based On Decentralized Sliding Mode Observers

With the development of technology, robotics has been improved step by step. It hasreached various fields in global economic development of all over the world, from capsuleand spacecraft to micro-robot and deep-sea explorer, as well as plays an important role ofmodern high technology. Therefore, it is hoped that the robots can undertake morecomplex work, and adapt to the unknown environments quickly. However, due to thelimitations of their own mechanical structure, the traditional robots can not complete alltasks. Luckily, reconfigurable robot is consist with a set of joints and links with differentsize and performance, which could be interchanged, added or removed to reconfigure arequired robot according to the special tasks. The reconfiguration includes not only themechanical structure reconstruction, but also the control system.The requirement of reliability and safety is increased with the widely application ofreconfigurable robot. In the sake of decrease the investment cost and operation risk, theresearch of fault diagnosis and tolerant control has attracted much attention. Due to thevarious tasks and complex working environments, it is very important to detect anddiagnosis of faults for the reconfigurable robot system actuators, sensors, as well as otherimportant components (joint module, link modules, etc.). It is hardly to imagine theeconomic loss and other consequences once the control system failure without beingtreated well. Thus, it has important significance in theory and wide application in thefuture.In this paper, the dynamic model and subsystem dynamic model of thereconfigurable robot system are given based on decentralized control cogitation, then forthe subsystem sensor and actuator fault of the reconfigurable robot, a decentralized slidingmode observer based fault diagnosis and tolerant control scheme is developed. The mainresearch contents in this paper are as follows:At first, the dynamic model of the reconfigurable robot is established based onNewton-Euler iterative algorithm automatically, then decomposes the entire system intoa set of coupled subsystems and gives the fault model with actuator and sensor fault.Secondly, a global sliding mode adaptive fuzzy controller, which is applied as thecontroller of the system under normal state, is designed based on the subsystem dynamicmodel of the reconfigurable robot. Then a fault diagnosis scheme is proposed for thesubsystem under sensor and actuator fault based on the decentralized sliding modeobserver. Introduce a new state by state variable extension which regards the sensor faultas pseudo-actuator fault, and for this total actuator fault, compares the residual signalobtained from the output of the nonlinear observer and the actual output with the given threshold. When the output residual signal is smaller than the threshold, the system runsnormally, otherwise, the fault of sensors or actuators of the system occurs. After that, thescheme which isolates and estimates the faults online based on the multi sliding modeobserver technique. The neural network is adopted to approximate the uncertainty andinterconnection among the subsystems of the observer model, and compensates theestimation error adaptively, which guarantees the robustness of the fault diagnosis system.Thirdly, an active decentralized fault-tolerant control method based on the bicyclicintegral sliding mode contains position loop and velocity loop is presented by utilizing thefault diagnosis and measurable joint velocity for the reconfigurable robot system withactuator and sensor fault. When the position fault occurs, the integral signal of thevelocity sensor as the position loop feedback value is adopted to instead the fault signal,and the RBF neural network is adopted to compensate the actuator fault to realize the faulttolerant control and to guarantee the stability and accuracy after the fault occur. When thejoint velocity signal is immeasurable, combine adaptive control and nonsingular terminalsliding mode, isolate the system fault to improve the convergence and reach the balancepoint in limited time, realizes the fault tolerant control and avoids the singularphenomenon appears in traditional terminal sliding mode control.Finally, numerical simulation is studied through3-DOF reconfigurable robot withdifferent configurations, and the results show the effectiveness of the proposed schemes.A summary of this work is given, and look forward to the future work to be studied.