Learning local trajectories for high precision robotic tasks : application to KUKA LBR iiwa Cartesian positioning

摘要

To ease the development of robot learning in industry, two conditions need tobe fulfilled. Manipulators must be able to learn high accuracy and precisiontasks while being safe for workers in the factory. In this paper, we extendpreviously submitted work which consists in rapid learning of local highaccuracy behaviors. By exploration and regression, linear and quadratic modelsare learnt for respectively the dynamics and cost function. Iterative LinearQuadratic Gaussian Regulator combined with cost quadratic regression canconverge rapidly in the final stages towards high accuracy behavior as the costfunction is modelled quite precisely. In this paper, both a different costfunction and a second order improvement method are implemented within thisframework. We also propose an analysis of the algorithm parameters throughsimulation for a positioning task. Finally, an experimental validation on aKUKA LBR iiwa robot is carried out. This collaborative robot manipulator can beeasily programmed into safety mode, which makes it qualified for the secondindustry constraint stated above.