An experimental study on improvement of weaving trajectories of welding robots by a learning scheme

摘要

Parameters of a weaving trajectory consisting of amplitudes and phases are learned so that an evaluating function of error which is computed from accelerometer data is minimized. Two methods of learning are applied to robot manipulators: a basic method with high stability and a high speed algorithm using a model. Experimental results showed that learning methods can successfully improve weaving trajectories. The second method considerably improved the learning efficiency to only less than 30 times of repetitions of the learning process at most. A good weaving motion along welding lines was realized without any sensor feedback by interpolating the learned parameter values.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>