Learning control for similar robot motions

In this paper, we propose a novel scheme for governing similar robot motions by using learning mechanisms. Most learning schemes need to repeat the learning process each time a new trajectory is encountered. The main reason for this deficiency is that the learning space for executing general motions of multi-joint robot manipulators is too large. To reduce the complexity in learning, we first classify robot motions according to their similarity. A new learning structure, which is motivated by the concept of a motor program, is then used to learn a class of motions. The proposed structure consists mainly of a fuzzy system and a CMAC-type neural network. The fuzzy system is used for learning of the samples in a class of motions. The CMAC-type neural network is used to generalize the parameters of the fuzzy system, which are appropriate for the control of the sampled motions, to deal with the whole class of motions. The learning process is performed only once and the learning effort is dramatically reduced for a wide range of robot motions.