The article discusses SICNav, a method for safe and interactive crowd navigation for robots. The system uses Model Predictive Control (MPC) to predict and react to human movements, avoiding collisions. It models each human in the crowd following an Optimal Reciprocal Collision Avoidance (ORCA) scheme. This model is embedded as a constraint in the robot’s local planner, leading to a bilevel non-linear MPC optimization problem. The proposed method can influence pedestrian motion while ensuring safety constraints in a single-robot multi-human environment. The performance of SICNav is analyzed in a simulation environment.
Publication date: 18 Oct 2023
Project Page: No link provided
Paper: https://arxiv.org/pdf/2310.10982
