Improving Traffic Flow in Lane-free Intersections via Decentralized Sampling-based Dynamic Trajectory Refinement

Lane-free traffic is a novel traffic control strategy in which vehicular movement is no longer restricted to fixed lanes and can occupy any lateral position within road boundaries. As such, portions of classical road infrastructure can be reallocated to other uses, including pedestrian, cycling, or transit infrastructure, while maintaining traffic capacity through the precise control capabilities of autonomous vehicles. Intersection control, however, remains particularly challenging due to frequent vehicle conflicts. Prior work predominantly relies on centralized approaches such as time-slot reservation or rule-based motion primitives, which may suffer from limited resolution of the search space or may fail to capture dynamic path-velocity interdependencies. To address these challenges, we propose a decentralized, sampling-based dynamic trajectory refinement framework. Under this paradigm, each vehicle refines its trajectory iteratively using a local, adaptive strategy that efficiently explores the search space for evasive maneuvers. By localizing refinement to the conflicting segment, our method supports trajectory continuity and reduces disruptions to overall traffic flow. The adaptive sampling further accelerates conflict resolution and helps prevent delay cascades under high demand. Our approach is benchmarked against established baselines, including fixed-time signal control, SlotIIC, and the FERSTT rule-based strategy. Simulation results demonstrate that our method could improve vehicular throughput by reducing vehicle conflicts and maintain computational efficiency as the traffic enters more saturated regimes.      «

Lane-free traffic is a novel traffic control strategy in which vehicular movement is no longer restricted to fixed lanes and can occupy any lateral position within road boundaries. As such, portions of classical road infrastructure can be reallocated to other uses, including pedestrian, cycling, or transit infrastructure, while maintaining traffic capacity through the precise control capabilities of autonomous vehicles. Intersection control, however, remains particularly challenging due to frequ...     »