基於車輛移動模式之即時車流壅塞偵測

城市道路塞車現象在全世界來說都是一個很嚴重的問題。現存的道路塞車狀態偵測方法大都要求收集許多來自同一條道路上的車輛的資訊以準確判斷出此道路的交通狀況，這樣對系統的市場滲透率要求非常高，實際操作中往往不可行。我們通過觀察發現，汽車在塞車和不塞車情況下的行為是有明顯區別的，因此我們提出一種新的方法，通過分析汽車的行為反推出其所在道路的交通狀況。這樣的好處是我們不需要預先知道任何資訊，也沒有任何基礎設施建設的成本，只需要利用安置在車內的智能手機，即可判斷出汽車所行駛道路交通狀況，因此在市場滲透率很低的情況下仍然可以部署。為此，我們使用了機器學習的方法，將汽車行駛的運動軌跡作為特徵值來對道路狀況進行分類。在建模階段，我們使用一種高精度的鐳射光雷達（LIDAR）收集真實世界中車輛行駛軌跡，以確保獲取數據的準確性和真實性。我們根據汽車行駛的道路狀況將收集到的汽車軌跡標記成“塞車”和“不塞車”兩個狀態。在實測階段，我們使用KNN 算法，將汽車軌跡和我們收集到汽車軌跡進行對比，用最接近的那條軌跡的標記來標記當前道路塞車情況。為了處理汽車軌跡不同長度的問題，我們使用了動態時間規整（dynamic time warping）的方法計算兩條軌跡之間的距離。實驗數據顯示，我們的系統在僅使用一輛汽車資訊的情況下可以得到接近90% 的準確率

關鍵字

塞車偵測；車聯網；機器學習

並列摘要

Traffic congestion in urban areas is a severe problem in many cities around the world. Existing solutions require data from a considerably large number of vehicles on the same road to accurately detect traffic congestion of a particular road. We notice that the behaviors of vehicles in congested and noncongested states are discriminable, thus we present a novel approach to detect the traffic states using motion patterns of an individual vehicle in real-time without any other preliminary knowledge. Specifically, only a smart phone is needed to predict the traffic state in our system. The biggest advantage of such an approach is that the system can function properly even if there are only a smaller number of vehicles equipped with the system, which is usually the case at the early stage of the deployment of a vehicle-to-vehicle (V2V) network or a large-scale intelligent transportation system. In our solution, machine learning mechanisms are utilized to classify the traffic state by extracting the motion behaviors of a vehicle. Our model development utilizes highly accurate vehicle traces collected at several real-world intersections with LIDAR. In implementation, we use KNN algorithm to determine the traffic state by comparing the real-world GPS trace with our collected traces. Since the length of traces will be different according to different road segments, we utilize dynamic time warping (DTW) mechanism to calculate the distance. Evaluation results show that, with only data from an individual vehicle, our classification model can achieve a promising 90% precision

並列關鍵字

Congestion Detection ； Machine Learning ； Vehicular Network

參考文獻

[11] J. Wedel, B. Schunemann, and I. Radusch, “V2x-based traffic congestion recognition

[17] J. Park, Z. Chen, L. Kiliaris, M. Kuang, M. Masrur, A. Phillips, and Y. Murphey,

[4] L. Li, L. Chen, X. Huang, and J. Huang, “A traffic congestion estimation approach

[3] P. Desai, S. Loke, A. Desai, and J. Singh, “Caravan: Congestion avoidance and

[2] G. Marfia and M. Roccetti, “Vehicular congestion detection and short-term forecasting:

國際替代計量

基於車輛移動模式之即時車流壅塞偵測

全文下載

主題瀏覽