Iterative Learning Control of Hard Constrained Robotic Manipulators

This paper considers the hard constraints in the movement of the joints of robotic manipulators executing repetitive tasks in the presence of measurement noise in the dynamic model. In our research work we have established that subject to these conditions the Iterative Learning Control (ILC) is one of the best methods performing such tasks with high precision. The existing applications of ILC don’t take in consideration the inherent nonlinearity of the dynamic model of a robotic manipulator neither the hard constraints in state space. This paper presents a robust and convergent Constrained Output ILC (COILC) method for non-linear hard constrained systems like industrial robots. Unlike known ILC methods the COILC employs a novel algorithm to cancel the currently executing iteration before the occurrence of a violation in any of the state space hard constraints. This way COILC resolves both the hard constraints in the state space and the transient growth problem, which is a major obstacle in applying ILC to non-linear systems. The convergence and the performance of the proposed numerical procedure are experimentally evaluated through computer simulation of the well-known SCARA-type robot TT-3000. The obtained results are discussed and demonstrate that COILC is suitable for solving hard constrained state space problems of non-linear systems in robotics with certain inaccuracies in the dynamic model.