Decentralized Navigation of Multiple Robots Based on Event-Triggered NMPC and NH-ORCA

This paper presents a decentralized navigation strategy for multiple differential-drive robots in densely populated environments. The strategy employs the Nonholonomic Optimal Reciprocal Collision Avoidance (NH-ORCA) algorithm to generate a collision-free velocity set for robots based on locally collected sensor data. Within this velocity set, the Nonlinear Model Predictive Control (NMPC) based planning problem is formulated and then solved to generate collision-free local trajectories for each robot. Additionally, this paper proposes an event-triggered mechanism to reduce the computational load of solving NMPC, enabling the algorithm to operate in real-time. The number of prediction steps of NMPC is adjusted according to the robot's velocity, so as to mitigate the loss of optimality of the event-triggering scheme. Simulation results show the advantages of the proposed decentralized navigation method.