以混合車流模型為基礎之最佳路網避障路徑規劃

Abstract

因應社會與科技的進步以及都市人口的快速增長與集中所造成的都市交通問題，各國開始進行智慧交通系統(ITS)的整合規劃。透過政策的制定、電子化交通系統設備與自主駕駛車輛的開發，期望能藉由中央管理之高效率的智慧交通路網達到安全又環保。本研究以細胞自動機(CA)為基礎的車流模型，建構符合台灣都市交通狀況的多線道汽機車混合車流擬真模型。利用此模型探討障礙對車流的影響，找出可以改善車流狀態車輛最佳避障策略。單路避障方法至都市路網模型，透過以道路權重為依據的Dijkstra演算法找出路網避障的最佳策略，進而提升整體路網的行車效率。最後經由模擬的得出，在單路避障的最佳化結果可以使到路車流量獲得改善，車輛密度在30至60%的情況下，可以大約提升20至40%的車流量。而在路網避障的最佳路徑規劃，車輛在為了避開障礙而選擇花費時間較少之路徑行駛，雖會延長移動距離，在不考慮因路徑改變導致道路車輛密度增加的話，特定狀況可節省大約60%旅行時間，若考慮車流密度變化則節省約4%的時間。

Urban traffic has faced a much greater challenge with the increase of vehicle population as society and technology advance. Modern intelligence transportation system (ITS) combines state-of-the-art communication and sensing techniques to provide a more efficient, more economic, and safer on-road environment. This research investigates the advantages of a centralized ITS management strategy with autonomous vehicle and develop a method to provide efficient traffic operation of mixed fleets when on-road obstacles are present. We use a refined cellular automata traffic model to better capture the characteristics of multi-lane mixed-fleet driving behaviors, such as lane-splitting and overtaking, in Taipei. An optimization is formulated to obtain the best obstacle avoidance strategy to maintain high traffic flow in a single road. Results show that with our method, an obstacle that usually cause serious congestion can improve the traffic flow by 20% (for 30% vehicle density) to 40% (for 60% vehicle density). We further extend the single-lane obstacle avoidance strategy to a network with multiple roads. Dijkstra algorithm is used to search for the minimal traveling time in network while each road could potentially occupied by multiple obstacles. Compared with being stocked on the original driving path, our method can suggest a new path with 60% less travel time. Our proposed method has the potential to include traffic lights or complex driving conditions for more effective and practical ITS in the future.