Physically-Proximal Human-Robot Collaboration: Enhancing Safety and Efficiency Through Intent Prediction

摘要

*† Future space missions may be accomplished with greater efficiency through humanrobot collaboration, with robots expected to autonomously schedule and execute repetitive or easily-automated tasks. This paper explores issues associated with a human and robotic manipulator operating in a shared physical workspace. A multi-layer autonomous control system is presented that enables a robot to determine near-term intent of a human companion and translate that intent into manipulator motions. These motions must be planned and adapted to avoid collisions with and minimize distraction to the human as both manipulator and human accomplish their mission tasks. This paper presents results from initial experiments designed to assess the impact of a physically-proximal manipulator on human task performance and workload. Human tasks are recognized upon initiation and translated to constraints on manipulator motion to avoid blocking the humans’s expected path to a target or visual gaze to a computer screen used to solve a sequence of simple cognitive (math) tasks. Preliminary results indicate the robot is able to accomplish tasks with minimal impact on a human companion’s performance or workload. Future work is planned to mature the robotic system’s perceptual and prediction capabilities and to complete a more comprehensive set of human-robot collaboration experiments.