Visual-spatial attention as a comfort measure in human-robot collaborative tasks

Abstract In this paper, we propose a new formulation to consider visual–spatial attention in order to improve the comfort of a human standing or operating near a collaborative robot. This formulation is based on the principle of having the robot’s end-effector in the human visual–spatial attention as much as possible. The integration of this new constraint into the Inverse Kinematics (IK) problem is thoroughly studied and efficient solutions are proposed. Moreover, to allow the robot to react rapidly in the case of unforeseen events, adding the manipulability index to the IK problem is also studied and its impact is analyzed. The proposed method is then extensively tested in simulation and verified on the real Baxter research robot, these experiments pointed out the method efficiency to improve the task visibility while avoiding self-occlusion and singular configurations. Moreover, real experiments revealed the method robustness to uncertainties such as imprecision of detecting human gaze direction.