Human-robot co-manipulation of extended objects: Data-driven models and\n control from analysis of human-human dyads

Abstract

Human teams are able to easily perform collaborative manipulation tasks.\nHowever, for a robot and human to simultaneously manipulate an extended object\nis a difficult task using existing methods from the literature. Our approach in\nthis paper is to use data from human-human dyad experiments to determine motion\nintent which we use for a physical human-robot co-manipulation task. We first\npresent and analyze data from human-human dyads performing co-manipulation\ntasks. We show that our human-human dyad data has interesting trends including\nthat interaction forces are non-negligible compared to the force required to\naccelerate an object and that the beginning of a lateral movement is\ncharacterized by distinct torque triggers from the leader of the dyad. We also\nexamine different metrics to quantify performance of different dyads. We also\ndevelop a deep neural network based on motion data from human-human trials to\npredict human intent based on past motion. We then show how force and motion\ndata can be used as a basis for robot control in a human-robot dyad. Finally,\nwe compare the performance of two controllers for human-robot co-manipulation\nto human-human dyad performance.\n