基於動態 QoS 映射演算法與 QoS 感知整形的低延遲遠 端駕駛系統

Abstract

本研究旨在實現低延遲且具備可靠度的遠端駕駛系統，並在 5G 和時間敏感網路 (TSN) 整合的系統架構下提出了一種新穎的系統架構來動態調整服務品質 (QoS) 映射並利用 QoS 感知流量整形器，確保在多車同時進行串流的無線環境中維持穩定的傳輸表現。單一輛車內不同類型的訊息（例如控制指令、狀態回報、攝影機影像資料以及點雲數據），乃至於多車之間的通信，均可能因其高頻率與大量數據傳輸的需求，對網路資源造成極大的壓力。當這些訊息競爭有限的網路頻寬時，可能引發網路壅塞現象，進一步導致傳輸延遲的顯著增加。此類延遲問題對延遲敏感的應用，特別是遠端駕駛或其他車輛即時應用，將產生嚴重影響，甚至威脅系統的穩定性與安全性。因此，如何有效管理車內及車間訊息的傳輸資源，並確保低延遲和高可靠性，成為智慧車輛通訊系統中亟需解決的關鍵挑戰。本研究透過將動態頻寬分配和自適應流量控制相結合，系統可以優化資源使用並最大限度地減少控制訊號等時間關鍵訊息的延遲。使用 OMNeT++ 模擬器進行的實驗驗證表明，所提出的框架有效減少了端到端延遲並提高了訊息即時性，即使在高網路負載下也能保持運作穩定性。

This work addresses the challenges of achieving low-latency and reliable communication in remote driving systems, particularly under the constraints of 5G and Time-Sensitive Networking (TSN) integration. A significant challenge arises within vehicles where internal applications compete for limited computation and communication resources, leading to potential bottlenecks. Additionally, in multi-vehicle systems, concurrent resource demands among vehicles exacerbate congestion, further degrading their performance. To mitigate these issues, this work proposes a traffic shaping mechanism that dynamically adjusts Quality of Service (QoS) mapping and employs a QoS-aware shaper mechanism. By leveraging dynamic bandwidth allocation, the system ensures the timely transmission of time-sensitive messages under various network loads. Experimental validation using the OMNeT++ simulator demonstrated that the proposed framework effectively shortens the end-to-end delay for non time-sensitive messages and enhances message timeliness for time-sensitive messages. The system maintains operational stability by balancing bandwidth efficiency and QoS reliability, achieving high-confidence, low-latency transmission even under heavy-load conditions. This work provides a robust approach to addressing resource contention challenges in latency-sensitive remote driving applications.