分散式路口管理中車對車通訊協定設計之分析

Abstract

隨著聯網與自駕車輛的快速發展，交叉路口的高效與安全管理需求日益增加。雖然集中式系統可透過車輛與基礎設施間的通訊（V2I）提供穩定的協調機制，但主要以車對車通訊（V2V）為基礎的分散式方法展現了更高的適應性與較低的基礎設施成本。然而，分散式方法也常因頻繁廣播帶來高通訊負擔。 本研究提出一種類似防火牆的V2V協定框架，透過動態建立訊息過濾規則以降低通訊負擔，同時維持車輛的情境感知能力。我們進一步設計了一個分散式交叉路口管理框架，並對三個提出的協定與兩個基準方法進行比較與評估。實驗結果顯示，部分具備過濾功能的協定在效率上可與不做任何訊息過濾的基準測試相當，顯示許多訊息在該框架下是冗餘的。採用更細緻的衝突區域劃分時，即使在高交通密度情況下，系統仍能維持可接受的低延遲。此外，系統在高達55%的訊息遺失率下仍展現出良好的穩健性。這些結果凸顯了我們所提出的框架在基礎設施有限與通訊干擾等真實環境中，具備實際部署的潛力。

The growing presence of connected and autonomous vehicles has accelerated the need for efficient and safe intersection management. While centralized systems offer reliable coordination through vehicle-to-infrastructure communication. In contrast, distributed approaches based on vehicle-to-vehicle communication offer greater adaptability and reduced infrastructure costs, but often incur high message overhead due to frequent broadcasting. This work proposes a firewall-style Vehicle-to-Vehicle (V2V) protocol framework that dynamically constructs message filtering rules to reduce communication overhead while preserving situational awareness. A distributed intersection management framework is developed to evaluate three proposed protocols against two benchmarks. Experimental results demonstrate that some protocols with filtering can match the efficiency of non-filtering baselines, highlighting that many messages are redundant. The system achieves an acceptable low delay even under high traffic density when finer conflict zone granularity is applied. It demonstrates robustness against message loss rates of up to 55\%. These findings underscore the potential of the proposed framework for deployment in real-world environments with limited infrastructure and communication interference.