Real-time safe motion generation through dynamical system modulation with multiple depth sensors

Abstract

Robots will be a part of our society in the future. They are expected to interact and collaborate with humans in daily life and production, which requires robots capable of adapting dynamical scenarios in real time. In this paper, we propose a real-time approach to generate safe motion of a robot when interacting with a human. Based on the concepts of dynamical system (DS) and artificial potential field (APF), the robotic system and the sensed obstacles are modeled as a DS by computing the minimum distance from the control point of a robot to the surfaces of obstacles, and then modulated to avoid collisions. The modulation is parameterized so that the safety margin could be dynamically modified according to specific situations, such as when encountering big uncertainty of the obstacle location. In order to address potential safety risk in case of occlusion occurs with only one sensor, two depth sensors are configured to monitor the entire shared workspace from different view angle in our approach. Experiments with a UR5 robot interacting with a human under the surveillance of two Xtion depth sensors are conducted to showcase the effectiveness of our proposed method.