The Braess paradox is a counter-intuitive phenomenon whereby adding roads to a network results in higher travel time at equilibrium. In this paper we present an algorithm to detect the occurrence of this paradox in real-world networks with the help of an improved graph representation accounting for queues. The addition of queues to the network representation enables a closer match with real data. Moreover, we search for routes causing this phenomenon ("Braess routes") rather than links, and advocate removing such routes virtually from navigation systems so that the associated links can continue to serve other routes. Our algorithm relies on a convex optimization problem utilizing Beckmann potentials for road links as well as queues, and results in a route reconfiguration with reduced delay. We assume the availability of historical data to build the optimization model. We also assume the existence of a centralized navigation system to manage the routing options and remove the Braess routes. The theoretical solution demonstrates up to 12% delay reduction in a network from Montgomery County, Maryland. We validate the improvement with simulations.
翻译:Braess 悖论是一种反直觉现象,通过将道路加到网络上,可以实现平衡的较高旅行时间。在本文中,我们提出了一个算法,在改进队列的图表解析功能的帮助下,在现实世界网络中发现这一悖论的发生。在网络代表中增加队列,可以更接近真实数据。此外,我们搜索造成这一现象的路线(“Braess路线”)而不是连接,并主张从导航系统中几乎删除这些路线,以便相关连接能够继续用于其他路线。我们的算法依赖于利用Beckmann潜力进行道路连接和排队的螺旋式优化问题,从而在缩短的时间内重新配置路线。我们假设历史数据的可用性数据来构建优化模式。我们还假设存在一个集中的导航系统来管理路线选项,并拆除Braess路线。理论解决方案显示,来自马里兰州蒙哥马利州的网络有12%的延迟。我们用模拟来验证改进情况。