Communication-free Distributed Control Algorithm for autonomous vehicles at intersections

Recently published:

• Alireza Soltani, David M. Levinson, Mohsen Ramezani (2025) Communication-free Distributed Control Algorithm for autonomous vehicles at intersections. Transportation Research Part C: Emerging Technologies Volume 180, November 2025, 105309 [doi]

ABSTRACT

This paper introduces a novel approach for managing autonomous vehicles at signal-free intersections through a Communication-free Distributed Control Algorithm (CfDCA). Unlike centralized systems or communication-based decentralized methods, CfDCA relies solely on onboard sensors and in-vehicle decision-making to ensure efficient and collision-free navigation. The algorithm formulates intersection management as a distributed optimization problem with demonstrated safety logics and robustness to measurement errors. The algorithm combines a dynamic resource acquisition graph with a refined priority function and an adaptive tolerance mechanism to ensure efficient performance under varying traffic conditions. A stochastic tie-breaking mechanism is proposed to handle rare cases of identical priorities, while deadlock prevention is guaranteed through strict priority ordering. Simulation experiments demonstrate that CfDCA reduces average delay and queue length and is able to achieve throughput higher than actuated signalized intersections and outperforms a first-come-first-served baseline in delay reduction. Additionally, the algorithm’s distributed design offers scalability and eliminates dependency on communication infrastructure.

Fig. 1. (a) Predefined paths and 16 conflict zones in yellow. The shaded areas with red demonstrate the spatial extent within which an AV in movement 1 monitors and evaluates other vehicles for conflict resolution. (b) Sizes and distance-based configuration of the intersection. The Line of Consideration (LC) defines the area within which AVs actively consider other vehicles for conflict zone acquisition. The Line of Detection (LD) defines the area where AVs start detecting other movements and include vehicles in the graph. AVs begin evaluating potential conflicts and requirements upon entering the LD. The colored shading illustrates the position-based vehicle sets: green areas represent the active set, blue areas represent the buffer set, and their combination forms the observation set.