Dynamic path planning in human-shared environments for low-resource mobile agents

Mobile agents are widely employed in industrial applications. However, in an environment where humans and robots coexist, the robot should be able to navigate autonomously while safely avoiding moving obstacles, especially human operators. In this paper, we propose a dynamic path planner providing an additional costmap layer in which the area around the detected human is inflated by a Gaussian cost. The latter is proportional to the obstacle's speed and orientation, leading to a safer avoidance behaviour during navigation. The algorithm, suitable for low-resource mobile agents, has been developed in ROS1 and then experimentally validated on the Locobot mobile manipulator in a laboratory environment.