Safe Robot Navigation in a Crowd Combining NMPC and Control Barrier Functions

We propose a sensor-based scheme for safe robot navigation in a crowd of moving humans. It consists of two modules, i.e., the crowd prediction and motion generation module, which run sequentially during every sampling interval. Using information acquired online by an on-board sensor, the crowd prediction module foresees the future motion of the humans in the robot surroundings. Based on such prediction, the motion generation module produces feasible commands to safely drive the robot among the humans by combining a nonlinear Model Predictive Control (NMPC) algorithm with collision avoidance constraints formulated via discrete-time Control Barrier Functions (CBFs). We show the effectiveness of the proposed approach via simulations obtained in CoppeliaSim on the Pioneer 3-DX mobile robot in scenarios of different complexity.