A Distributed Consensus Algorithm for Autonomous Vehicles Deciding Entering Orders on Intesections without Traffic Signals

2025-07-04

AI Summary

Coordination of right-of-way decisions for mixed traffic—comprising connected and autonomous vehicles (CAVs) and human-driven vehicles (HVs)—at unsignalized intersections remains challenging due to asynchronous communication, vehicle failures, and real-time constraints. Method: This paper proposes a distributed voting algorithm inspired by the Raft consensus protocol. It features a lightweight candidate and leader election process, enforces a minimum quorum to ensure safety and liveness under asynchrony and partial failures, employs gRPC for inter-vehicle communication, and integrates computer vision to estimate vehicle arrival order—with license-plate lexicographic ordering as a fallback during network timeouts. Contribution/Results: Evaluated on a canonical four-leg, two-lane intersection, the approach achieves average consensus latency of 30–40 ms. It demonstrates robustness across varying CAV penetration rates, intersection scales, and vision-induced delays. To our knowledge, this is the first systematic application of distributed consensus paradigms to cooperative control in mixed-autonomy intersections, balancing real-time performance, reliability, and scalability.

Technology Category

• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems
• Humans and AI: Voting
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization

Application Category

• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Security and Privacy: Large-scale security measurements

Abstract

We propose a methodology for connected autonomous vehicles (CAVs) to determine their passing priority at unsignalized intersections where they coexist with human-driven vehicles (HVs). Assuming that CAVs can perceive the entry order of surrounding vehicles using computer vision technology and are capable of avoiding collisions, we introduce a voting-based distributed consensus algorithm inspired by Raft to resolve tie-breaking among simultaneously arriving CAVs. The algorithm is structured around the candidate and leader election processes and incorporates a minimal consensus quorum to ensure both safety and liveness among CAVs under typical asynchronous communication conditions. Assuming CAVs to be SAE (Society of Automotive Engineers) Level-4 or higher autonomous vehicles, we implemented the proposed distributed consensus algorithm using gRPC. By adjusting variables such as the CAV-to-HV ratio, intersection scale, and the processing time of computer vision modules, we demonstrated that stable consensus can be achieved even under mixed-traffic conditions involving HVs without adequate functionalities to interact with CAVs. Experimental results show that the proposed algorithm reached consensus at a typical unsignalized four-way, two-lane intersection in approximately 30-40 ms on average. A secondary vision-based system is employed to complete the crossing priorities based on the recognized lexicographical order of the license plate numbers in case the consensus procedure times out on an unreliable vehicle-to-vehicle communication network. The significance of this study lies in its ability to improve traffic flow at unsignalized intersections by enabling rapid determination of passing priority through distributed consensus even under mixed traffic with faulty vehicles.

Problem

• Determining passing priority for autonomous vehicles at unsignalized intersections.
• Resolving tie-breaking among simultaneously arriving autonomous vehicles.
• Ensuring safety and liveness in mixed traffic with human-driven vehicles.

Innovation

• Voting-based distributed consensus algorithm for CAVs
• Uses gRPC for implementing the consensus algorithm
• Secondary vision system for priority if consensus fails