Proactive trajectory deconfliction for human-robot collaboration: a predictive velocity-space approach using dynamic safety ellipsoids

Abstract Human–robot collaboration in manufacturing and service tasks is limited by the conflicting requirements of uninterrupted task execution and reliable worker safety. Current safety frameworks rely on reactive procedures such as emergency stops, which disrupt workflow and reduce the quality of interaction. A novel trajectory deconfliction scheme is presented through the use of a Dynamic Safety Ellipsoid that represents predicted human movement, incorporating uncertainty in position. The ellipsoid creates a predictive boundary around human motion, allowing the robot to modify its trajectory in real time rather than halting operations. Collision prevention is achieved through velocity-space control, where a repulsive velocity vector is introduced within the manipulator’s null space. The proposed method maintains the main task integrity and prevents the commanded trajectory from being discarded. A Kalman filter is embedded for human motion estimation, and this addition improves the prediction accuracy of motion trajectories. Simulation validation is carried out through a 7 degree of freedom robotic manipulator model implemented in the Robotics Toolbox environment. Comparative evaluations demonstrate a 63.8% reduction in task execution time when measured against reactive stopping and replanning strategies, and the safety buffer is preserved at an equivalent level. The method further yields smoother motion profiles, which are confirmed by the reduced value of integrated jerk. These results confirm that the Safety Ellipsoid strategy supports safer interaction, faster execution, and smoother operation. This approach enables continuous functioning without unnecessary halts, providing a reliable pathway for introducing collaborative robots into uncertain and dynamic working environments.