Growth in urban population, travel, and motorization continue to cause an increased need for urban projects to expand road capacity. Unfortunately, these projects also cause travel delays, emissions, driver frustration, and other road user adversities. To alleviate these ills, road agencies often face two work zone design choices: close the road fully and re‐reroute traffic or implement partial closure. Both options have significant implications for peri‐construction road capacity, traveler costs, and the project duration and cost. This study presents a decision‐making methodology to facilitate the choice between full road closure and partial closure. The presented decision‐making methodology is a bi‐level optimization problem: at the upper level, the road agency seeks to optimally schedule road construction work to minimize net vehicle emissions and road construction costs. The lower‐level of the problem captures two types of travelers’ route choice behaviors: rational travelers who minimize their travel time and path‐loyal travelers who do not change their routes from their pre‐construction routes. The bi‐level mixed integer nonlinear model is solved using a reinforcement learning‐based algorithm (the multi‐armed bandit‐guided particle swarm optimization [PSO] technique). The computational experiments suggest the superiority of the proposed algorithm, compared to the classic PSO algorithm in terms of solution quality. The numerical results suggest that if the percentage of path‐loyal travelers increases, the agency needs to invest more in road project construction to implement under partial closure to avoid a significant increase in vehicle emissions.

中文翻译：

---

考虑替代封闭类型的基于强化学习的城市道路项目调度方法


城市人口、出行和机动化的增长继续导致对城市项目扩大道路容量的需求增加。不幸的是，这些项目还会导致旅行延误、排放、驾驶员沮丧和其他道路使用者的逆境。为了缓解这些弊病，道路机构经常面临两个工作区设计选择：完全关闭道路并重新安排交通路线或实施部分封闭。这两种选择对施工周边道路容量、旅客成本以及项目持续时间和成本都有重大影响。本研究提出了一种决策方法，以促进在完全封闭和部分封闭之间进行选择。所提出的决策方法是一个双级优化问题：在上层，道路机构寻求以最佳方式安排道路建设工作，以最大限度地减少车辆净排放和道路建设成本。问题的较低级别捕获了两种类型的旅行者的路线选择行为：最小化旅行时间的理性旅行者和不改变施工前路线的路径忠诚旅行者。双级混合整数非线性模型使用基于强化学习的算法（多臂老虎机引导粒子群优化 [PSO] 技术）进行求解。计算实验表明，与经典的 PSO 算法相比，所提出的算法在求解质量方面具有优越性。数值结果表明，如果忠于路径的旅行者的百分比增加，该机构需要加大对道路项目建设的投资，以便在部分封闭的情况下实施，以避免车辆排放的显着增加。