Research On Iterative Learning Control Algorithm And Its Application In Manipulator

Iterative learning control (ILC) improves some control objects by successive iterativemodifying, which is particularly suitable for tracking the desired trajectory accuratelywithin a limited and constant time interval in a controlled object with a repetitive motion.Compared with other algorithms, ILC is not relied on the accurate mathematical models ofthe system but designed control input signals by making full use of previous control inputand errors. ILC do not need identifying parameters of the system as well as much priorknowledge. Consequently, ILC provides an effective method to the highly nonlinear,close-coupling and timevarying system such as robotic system. Based on the existingliteratures, new controllers are designed and simulations are performed by this paper tosolve different problems. The specific work as following:Firstly, a non-causal ILC algorithm with variable index gain is presented aiming atthe tracking for manipulator systems with initial error. This algorithm eliminates thelimitation that the initial states are consistent with the desired states or the initial states arefixed per iteration. Compared with fixed learning gain, ILC with variable index gain notonly avoids choosing gain aimlessly but also significantly enhances the convergent speed.In this paper, the proposed non-causal algorithm uses error information of future time inprevious iteration, then the controller can offset future disturbances in advance. From theperspective of execution, the proposed scheme is easy to implement. Moreover, weanalyze the convergence of the new algorithm. Finally, applying the proposed algorithm inmanipulator trajectory tracking system, we present simulation results to illustrate theeffectiveness of the control scheme.Secondly, as ILC algorithm does not need precise mathematical models and hascertain robustness to the unmodeled system. So this paper presents a new ILC algorithmwith angle correction term, which is studied for a class of nonlinear system with uncertainstate disturbance and arbitrary initial error. The angle relationship of outputs vectors isused as a standard to estimate the quality of the control inputs, and ‘awarding orpunishing’ is used to estimate the changing trend of the algorithm. So the convergentspeed is improved greatly. Finally, a non-causal iterative learning algorithm with variable forgetting factor isproposed for a general class of nonlinear system. The introduced variable forgetting canadjust the control input effectively and do filtering in the iteration direction, thus itaccelerates the convergent speed and weakens the influence on the convergence which isgenerated from the indeterminacy. Compared with the conventional PD-type algorithm,this algorithm not only increases information content but also avoids using the derivativeof error signals. The new algorithm presented in the paper is used in the simulation test ofmanipulator trajectory tracking, and the result illustrates effectiveness and real-timeperformance.