COMPARISON AND COMBINATION OF LEARNING CONTROLLERS - COMPUTATIONAL ENHANCEMENT AND EXPERIMENTS

Five discrete-frequency linear learning-control laws are compared and experimentally tested, These include simple integral-control-based lea rning using a single learning gain, phase-cancellation learning contro l, a contraction-mapping learning-control law with monotonic decay of the error norm, and leaning controllers that invert the system model a nd the observer model. The inversion designs converge the fastest init ially, but phase cancellation with identification updates and the cont raction-mapping method with model updates have better stability robust ness properties. The learning control approaches are combined to obtai n the advantages of each, by using inversion methods for the first few repetitions, followed by a more robust method, It is demonstrated tha t the computation of the learning action can be made in the frequency domain using fast Fourier transform methods, with as much as 94% reduc tion in computation time, In experiments on a Robotics Research Corpor ation robot, the learning-control laws result in a reduced rms trackin g errors for all joints, by a factor of close to 3 orders of magnitude .