基於數位孿生之適應性號誌控制方法

Abstract

交通號誌控制對號誌化路口的運作效率及智慧城市的建設至關重要。然而，在臺灣，許多路口仍依賴固定時制的號誌控制方法，此方法無法因應隨時間變化且具隨機性的交通需求，容易導致車隊過長、路口壅塞及車輛延滯等問題。本研究提出一種結合數位孿生模型的自適應交通號誌控制方法。系統透過影像監控設備蒐集即時車流資料，並配合統計管制方法，在偵測到交通流量出現顯著變化時，即以最新資料建立車流到達的隨機模型。接著，運用數位孿生模型模擬不同號誌時制方案，評估其平均車輛延滯時間，最終即時更新號誌為效能最佳的設定。 基於上述模型，本研究構建一個整數規劃問題，在實務考量的限制下以最小化平均車輛延滯為目標。透過持續監測，系統能在到達模式轉變時重新求解最佳設定，提供比固定時制更具反應能力的替代方案，進而提升路口運作效率與駕駛體驗。此概念是智慧城市發展的重要基礎，藉由大規模、資料驅動的交通管理，減少壅塞與污染排放，同時改善路網可靠性。

Traffic signal control is pivotal to the efficiency of signalized intersections and to smart-city operations. In Taiwan, however, many intersections still rely on fixed-time plans that cannot adapt to time-varying, stochastic demand, leading to long queues, excessive delays, and congestion. This study proposes a digital twin-enabled adaptive control strategy that updates signal settings whenever material changes in traffic flow are detected. The system employs camera-based vehicle detection to monitor real-time traffic conditions, constructing a digital twin simulation environment that mirrors actual intersection operations and integrating a stochastic arrival model to evaluate delays under candidate signal settings. Based on these models, we formulate an integer-programming problem that minimizes average vehicle delay subject to operational constraints. Continuous monitoring enables online re-optimization as arrival patterns evolve, yielding a responsive alternative to fixed-time control and improving the efficiency of intersection operations and the driving experience. This capability is foundational to Smart City development, enabling scalable, data-driven traffic management that cuts congestion and emissions while improving network reliability.