Implementation of SARL* Algorithm for A Differential Drive Robot in a Gazebo Crowded Simulation Environment
Abstract
Because of the stochasticity in people's behaviors, autonomous navigation in crowded environments is critical and challenging for both the robot and people evolving around. This paper deals with the implementation and effectiveness evaluation of the Socially Attentive Reinforcement Learning star algorithm, namely SARL*, which is an extended version of the state-of-theart socially compliant navigation algorithm SARL. It introduces a dynamic local goal resetting mechanism. The Simulations were conducted in the Robot Operating System (ROS) and the Gazebo simulator is used to test the human-aware navigation in different scenarios. Simulation results illustrate the efficiency of SARL* in terms of navigation around people in a socially acceptable manner. Nevertheless, it could not navigate efficiently when the goal position is located behind static or quasi-static obstacles.