Design And Application Of Iterative Learning Control Algorithms

Iterative learning control (ILC) makes a good use of repetitive information of aprocess to improve the control performance to obtain the desired output trajectories.Compared with the traditional control method, ILC can not only deal with the highuncertain relative degree in a simple way, but also acquires little priori knowledge and lesscomputation. What’s more, ILC is an algorithm which doesn’t depended on detailed modelof the controlled system and operate an optimizing input signal by iteration so that thesystem output can track the desired trajectory as closed as possible, and it is strongadaptive and easy to be implemented. The research on ILC is of great significance for suchnonlinear, close-coupling object as the robot.ILC algorithm in this paper is focus on robot control problem.First, rapid and high-precision iterative learning control (ILC) algorithm with twocompensations was proposed for robot trajectory tracking. The method picked upmany-sided information and compensated input and error respectively, it contributed tostrengthen the memory ability of system in the ILC process, what’s more, the output couldconverge to the desired output as quickly as possible and the miniature shakes and peaksin the error convergence can be eliminated effectively so that the trajectory tracking couldbe faster and be more precise, and there was some value to the high-precise robot’scontrol.Second, according to high-order relative degree systems need to be controlled byhigh-order differential learning laws. A discussion was presented on the iterative learningcontrol algorithm for the robot, which was an unknown relative degree system. First, adummy system with unit relative degree was presented by constructing a dummy model.Because of the special structure of the dummy model, there was an inevitable error intracking, a scheme was introduced to eliminate the error by reconstructing the desiredoutput.Finally, according to the indirect ILC which existed the uncertainties, disturbancesand noises, some work was launched as follows: A predictive control was used in the outer loop to update related parameter settings or the system would not be stable, so a sufficientcondition was introduced to choose a suitable predictive coefficient matrix for realizingthe robust asymptotically stability; Because the prediction was right or wrong there was nostandard to judge, a judgment section was added to the predictive section and thejudgment section could not only realize a function of judgment but also realized a functionof correction by “rewarding-punishing” the original predictive control law so that it couldget a good tracking performance.