Safe Motion Planning for Human-Robot Interaction

Jae Sung Park, Chonhyon Park and Dinesh Manocha
University of North Carolina at Chapel Hill



We present a motion planning algorithm to compute collision-free and smooth trajectories for robots cooperating with humans in a shared workspace. Our approach uses offline learning of human actions and their temporal coherence to predict the human actions at runtime. This data is used by an intention-aware motion planning algorithm that is used to compute a reliable trajectory based on these predicted actions. This representation is combined with an optimization-based trajectory computation algorithm that can handle dynamic, point-cloud representations of human obstacles. We highlight the performance of our planning algorithm in scenarios with a 7-DOF Fetch robot arm operating in a workspace with a human performing tasks. We demonstrate the benefits of our intention-aware planner in terms of computing safe trajectories.


Intention-Aware Motion Planning Using Learning Based Human Motion Prediction (Arxiv)
Arxiv Tech. Report, 2016

Algorithm Description

Incremental Trajectory Optimization for Motion Planning (ITOMP) is an optimization-based algorithm for motion planning in dynamic environments. Two key features of ITOMP are smoothness of planned robot trajectory and replanning capability responsive to changing environments near a robot. While environments around the robot are updated from sensor data in every time step, ITOMP repeatedly refines the trajectory to avoid collisions by interleaving motion planning and execution step.

Human motion prediction is an essential part for safe motion planning. Since human motion is fast and highly unpredictable, the robot system should be responsive to human motions enough to avoid future possible collisions which may happen. To make the motion planning system responsive enough to avoid collisions, we predict future human motions and consider the prediction human poses as obstacles in ITOMP. Our approach uses machine learning techniques to predict human motions; human motions are trained off-line and predicted in real time. In this way, the robot trajectory avoids before the human actually blocks robot's path, resulting in higher responsiveness and safety.


Safe Motion Planning with Human Motion Prediction Algorithm

Related Links

GAMMA Research Group
Motion Planning Research at GAMMA

Incremental Trajectory Optimization for Motion Planning (ITOMP)