A Subsumption Scheme for Emergent Collaboration of Self-Driving Vehicles in Intersections

This paper introduces a decentralized collaboration mechanism for self-driving vehicles at intersections, employing a subsumption scheme that re-uses individual agents’ motion planning. Agents exchange intended trajectories and adjust their decision-making based on received intentions. Agents’ action selection is carried out via competition of (many) motion plans. The results of this competition are emergent, safe behaviors. Evaluation with two and three vehicles at an orthogonal intersection showcases effective collaboration, where safety is ensured by reusing agents’ motion planning within the subsumption scheme.