利用真實路邊停車資料預測空位實作路邊停車導引系統

Roadside Parking Guidance System based on Availability Prediction Model with Real-World Data

吳浩平(Hao-Ping Wu)

指導教授 ： 蔡欣穆

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2021年)

https://doi.org/10.6342/NTU202104246

摘要

過去數十年，日益成長的都市人口帶給城市交通很大的挑戰。以停車為例，因為路邊車位的低成本與方便性，造成需求大增，而尋找車位的車輛在市區繞行造成交通壅塞，不但造成環境污染且浪費大量的時間及金錢，一套可以在都市環境中精準預測未來路邊車位的供需狀況系統成為大都市的迫切需要，例如台北。利用真實世界中的路邊車位歷史資訊，配合泊松過程模擬車輛進出行為，來預測未來停車格位供需情況，並利用SUMO模擬各項導引策略對於系統表現的影響。
本論文為實作都市路邊車位導引系統主要克服了以下三大挑戰。首先，將城市的路邊車位位置資訊整合並匯入車流模擬軟體，如此才能真實模擬並比較各導引策略的優劣。我們的模擬包含了1,557個路段，總共33,574個停車格位。再者，我們統計了2020年三月至2021年三月台北市所有路邊車位的車輛佔有資料並以泊松過程描述每個路段的車輛進出行為，計算出車輛駛入和駛離車位的頻率，並利用此資料預測未來車位的供需情況。最後，我們透過SUMO分析各策略對於整體使用者找尋車位的時間消長變化。在全部車輛遵循我們的導引下，與傳統貪婪的找尋車位方式相比，有超過百分之六十的車輛能減少找尋車位的總時間，其中有超過一半的車輛能減少超過十分鐘。因著我們系統的通用性，未來只要收集都市車位的歷史資料並將車位位置及都市路網匯入交通模擬軟體，便能對不同導引策略在不同都市進行評估，發展出適合該城市的都市路邊車位導引系統。

關鍵字

路邊停車 ； 停車導引 ； 泊松過程

並列摘要

In the last few decades, metropolitan population continued to grow, increasing urban traffic imposes remarkable challenges on the transportation system infrastructure. Take parking as example, since roadside parking is low-cost and convenient for driver, the demand for roadside parking exceeds the supply. There are many cars cruising for parking when they have already arrived their destination. Congestion, pollution, and parking issue are main problems for government. A parking space guidance system with predicting the parking demand changes is needed for metropolitan city like Taipei, Taiwan. By utilizing real world historical parking occupancy data and parking demand changes fitted by Poisson process, our system can adaptively adjust parking recommendations for users based on their current and predicted travel progress. In addition, SUMO simulation can evaluate the performance of our recommendation algorithm.
In this thesis, we conquer three main challenges. First, we implement a simulation framework and import real-world large-scale data in Taipei including 33,574 parking spaces distributed 1,557 road segments. Second, we collect historical parking occupancy data in Taipei from March 2020 to March 2021 and design our fitting algorithm by utilizing Poisson process to obtain the future parking demand changes for each road segments. Third, we create the Taipei city mobility simulation framework to evaluate the performance of our recommendation algorithm. We found that if all vehicles follow our guidance system, there are over 60% of vehicles can reduce their travel time comparing with greedy approach. Furthermore, there are over a half of vehicles have at least 10 minutes travel time improvement. Thanks to our framework's generic, one can evaluate the performance difference between different guidance strategies in metropolitan city by collecting historical parking occupancy data and importing city mobility, and thus everyone can develop their own parking guidance system according to the traffic characteristic.

並列關鍵字

on-street parking ； parking guidance ； Poisson process

參考文獻

[1] SEC ’17: Proceedings of the Second ACM/IEEE Symposium on Edge Computing, New York, NY, USA, 2017. Association for Computing Machinery.

[2] R. Arnott and E. Inci. An integrated model of downtown parking and traffic conges­ tion. Journal of Urban Economics, 60(3):418–442, 2006.

[3] P. Basu and T. D. C. Little. Networked parking spaces: architecture and applications. In Proceedings IEEE 56th Vehicular Technology Conference, volume 2, pages 1153– 1157 vol.2, 2002.

[4] M. Caliskan, A. Barthels, B. Scheuermann, and M. Mauve. Predicting parking lot occupancy in vehicular ad hoc networks. In 2007 IEEE 65th Vehicular Technology Conference ­ VTC2007­Spring, pages 277–281, 2007.

[5] M. Caliskan, D. Graupner, and M. Mauve. Decentralized discovery of free park­ ing places. In Proceedings of the 3rd International Workshop on Vehicular Ad Hoc Networks, VANET ’06, page 30–39, New York, NY, USA, 2006. Association for Computing Machinery.