一個指紋辨識定位演算法最佳取樣時間之研究

Abstract

智慧型運輸系統(Intelligent Transportation System, ITS)發展成功的關鍵就是取得正確且即時的交通資訊，以便分析與運用。近年有許多相關技術研究，尤其是利用流動車輛上的手機訊號資料(Cellular Floating Vehicle Data, CFVD)透過許多匿名的手機位置資料為樣本數與指紋辨識定位演算法(Fingerprint Positioning Algorithm, FPA)去取得精確的車輛速度和旅行時間來預測即時的交通訊息，但是針對車速回報之最佳取樣週期還沒有相關研究。其中，在取樣週期分析上，若週期太短則可能產生龐大的資料運算需求，造成伺服器負擔；若週期太長則可能造成取得車速回報的數量太少，而無法準確預測車速。因此，本研究將針對指紋辨識定位演算法的取樣周期進行分析及研究進而提出2個數學模式來取得最佳取樣週期。數值分析的結果顯示在取樣週期大於41.589秒後無用的資料量開始多於有用的資料量，效能是很差的，而小於41.589秒的取樣週期成本太高，效率是不好的。而實驗結果中顯示，當平均通話時間為60秒時，於本研究提出之方法可得最佳取樣週期為41.589秒，在此取樣週期下可產生足夠的車速資訊量且不會造成伺服器負荷過重，並且所得平均車速誤差率僅2.87%。因此，在智慧型運輸系統的建置上可採用本研究提出之成果，依各路段環境和通話行為建立最佳取樣週期，以提供即時車速資訊供駕駛決策參考。

Using cellular floating vehicle data has become an important technique to measure and forecast the real-time traffic information based on anonymously sampling the positions of mobile phones for intelligent transportation system (ITS). However, ITS server has a heavy load under higher sampling frequency, and traffic information cannot be provided immediately under longer sampling period. In this paper, two analytical models are proposed to analyze the optimal sampling period based on communication behavior, traffic conditions, and the two consecutive fingerprint positioning locations from the same call and estimating vehicle speed. In experiments, the result shows that the optimal sampling period is 41.589 seconds when the average call holding time is 60 seconds and the average speed error rate is only 2.87%. ITS can provide the correct and real-time speed information with lighter loads under the optimal sampling period. Therefore, the optimal sampling period of using fingerprint positioning algorithm is suitable to estimate speed information immediately for ITS.