ILC Design And Simulation Study For Two-link Robotic Manipulator Arm

Control problem of non-linear system is one of the topics which have been received more attention. It is well known that non-linear time-varying systems are very difficult to control and their controllers are very difficult to design even based on modern control theory. But for systems that track the same trajectory repetitively, such as robotic manipulators, iterative learning control (ILC) is a simple and effective approach to solve problems of such systems. ILC can improve the tracking accuracy between the real output and desired output over a fixed period of interval.In this paper, learning laws of three types of iterative learning control schemes, i.e. D-type, P-type and A-type ILC systems, are designed based on a two-link robotic manipulator arm model, respectively. Appropriate learning gains are chosen based on convergence analysis of each learning law to complete the design of controller. On the basis of the above research work, convergence of the three learning control schemes and robustness of system under the condition of each learning law are analyzed based on simulation. The aim is to study performance of three types of iterative learning control algorithms and their applications to a two-link robotic manipulator arm.The simulation results show that D-type and Anticipatory Sampled-data ILC systems satisfy the requirements of convergence, accuracy and robustness to dynamics disturbances, output measurement noises and errors in initial condition, and their convergence speed is fast enough. Furthermore, the effectiveness of the above two algorithms are verified. Compared with the above two ILCs, the convergence condition of P-type ILC is much stricter and its robustness is worse. Besides, convergence speed of P-type ILC is much slower than that of D-type ILC.Considering the real execution of a two-link robotic manipulator arm, D-type ILC can't achieve the effects of simulation results. The reason is that D-type ILC composes the learning law utilizing the derivative signals of the previous errors while there is no such problem that exists in P-type and A-type ILCs from the composition view of learning law. Thus D-type ILC may easily bring in noise and its effectiveness is reduced while P-type and A-type ILCs can restrain noise effectively. The effectiveness of A-type ILC is much better than that of D-type and P-type ILCs, because it has the anticipatory characteristics of the D-type ILC and the simplicity for implementation of P-type ILC.