A whole-body human-machine interface for dynamic bilateral teleoperation of humanoid robots

Abstract

Supported by the cumulative results of industrial robotics and legged locomotion research, humanoid robotics research has gained great strides in the past decade. The state-of-the-robot humanoid robots have demonstrated human-level dynamic motion capabilities in simple pre-programmed behaviors, such as running and jumping. However, most of these sophisticated and dynamic humanoid robots fall short of societies’ expectations when they attempt to perform seemingly ordinary but useful tasks in the real world, such as opening a door using the door handle, traversing uneven terrains, and interacting with humans. This is largely because these humanoid robots’ artificial brains lack the intelligence and intuition required for planning complex sequences of actions in the face of uncertainties. On the other hand, humans inherently possess such intelligence and intuition. Hence, humanoid robot teleoperation, i.e., controlling humanoid robots with human movements, promises to supply human planning intelligence to the robots, and appears to be a feasible intermediate step in the transition from physically capable humanoid robots to dynamically intelligent humanoid robots. Recent humanoid robot teleoperation results have also demonstrated this method’s potential, as teleoperated humanoid robots approached some autonomous humanoid robots’ superior dynamic performances with significantly simpler control algorithms. Thus, this work is dedicated to the development of a whole-body human-machine interface (HMI) as the medium through which a human and a robot communicate to achieve humanoid robot teleoperation. Specifically, the contribution of this work is the introduction of the HMI’s design and evaluations of the HMI’s performances regarding actuation force capability and sensing effectiveness. Furthermore, an experiment was conducted using the HMI to study physical human-robot interaction in arm teleoperation and compare the robot’s dynamic performances under joint space and task space teleoperation mappings.