An adaptive control approach based on dynamic movement primitives for human-robot handover

Abstract

Recent advancements on robotic technology expect a strong collaboration between humans and robots in the near future.Such collaboration relies on natural interaction between human and robot, demanding human-like behavior of the robot.A common scenario of human-robot collaboration is a human-robot handover.For this to be done seamlessly and in a human fashion, the robot has to anticipate the human agent's motion by predicting place and time of the handover on-line as soon as the human initiates the handover process.We consider the prediction problem as a model-based estimation problem where the point attractor and the timescale constant of the system are estimated on-line.Using Dynamic Movement Primitives as a parameterization of human motion, point attractor and timescale are successfully estimated on-line using different estimation laws based on adaptive control methods.Their stability is shown and their performance is evaluated using experimental data of human-human handovers.Thanks to the good prediction of the handover place, the presented algorithms are found to be applicable to improve human-robot collaboration.