physical Human-Robot interaction

This project proposes a method for imitating a nonlinear model predictive control law via deep neural networks, aimed at safely allowing a robot manipulator to share its workspace with a human operator, through the concept of "Speed and Separation Monitoring'' (SSM). The method devised in this project relies on a dataset-aggregation approach to explore the state space, and employs a safety filter to guarantee that the robot motion generated via deep neural networks still satisfies the safety constraints imposed by SSM. The proposed method is tested experimentally on a UR5 manipulator, and compared with related methodologies showing performance improvement.