- 學位論文
霧雲計算系統之高速公路旅行時間即時預測
Real-time Freeway Travel Time Prediction in a Fog-cloud Computing System
摘要
旅行時間預測在智慧運輸系統(ITS)中是一個重要議題，它可用於交通號誌控制、行程規劃、路線引導等應用。目前旅行時間預測的相關研究主要著重於在雲端環境中的預測準確性，但可能無法滿足即時的需求。為了實現即時且準確的旅行時間預測，我們提出了一個基於霧雲計算架構的高速公路旅行時間預測系統(TTP-FC)，它使用長短期記憶(LSTM)及漸進梯度回歸樹(GBRT)作為預測模型。根據從臺灣高速公路資料蒐集系統(TDCS)收集的數據，實驗結果顯示TTP-FC預測的平均絕對百分比誤差(MAPE)為2.145%，優於採用隨機森林模型的單一目的地預測(OTTP)的3.443%。此外，相較於僅在雲端環境執行，TTP-FC能降低26%的平均反應時間。
並列摘要
Travel time prediction is an important issue in Intelligent Transportation Systems (ITS), and it can be used for traffic control, route planning, and travel guidance. Existing studies on travel time prediction focus on prediction accuracy in cloud environments, which may not meet a real-time constraint. In order to achieve real-time as well as accurate travel time prediction, we propose a freeway Travel Time Prediction system based on a Fog-Cloud computing paradigm (TTP-FC), using a prediction model that combines the long short-term memory (LSTM) model with the gradient boosting regression tree (GBRT) model. Based on the data collected from the Traffic Data Collection System (TDCS) in Taiwan, evaluation results show that the average MAPE (mean absolute percent error) of the proposed TTP-FC is 2.145%, which is less than that (3.443%) of OTTP, a method based on random forests. In addition, the proposed TTP-FC can reduce the average response time by 26%, compared to that implemented in the cloud only.