Electric vehicles are a central component of future mobility systems as they promise to reduce local noxious and fine dust emissions and CO2 emissions, if fed by clean energy sources. However, the adoption of electric vehicles so far fell short of expectations despite significant governmental incentives. One reason for this slow adoption is the drivers' perceived range anxiety, especially for individually owned vehicles. Here, bad user-experiences, e.g., conventional cars blocking charging stations or inconsistent real-time availability data, manifest the drivers' range anxiety. Against this background, we study stochastic search algorithms, that can be readily deployed in today's navigation systems in order to minimize detours to reach an available charging station. We model such a search as a finite horizon Markov decision process and present a comprehensive framework that considers different problem variants, speed-up techniques, and three solution algorithms: an exact labeling algorithm, a heuristic labeling algorithm, and a rollout algorithm. Extensive numerical studies show that our algorithms significantly decrease the expected time to find a free charging station while increasing the solution quality robustness and the likelihood that a search is successful compared to myopic approaches.
翻译:电动车辆是未来机动系统的核心组成部分,因为它们承诺减少当地有毒和细灰尘排放和二氧化碳排放,如果由清洁能源提供的话。然而,尽管政府采取了大量奖励措施,但电动车辆的采用迄今没有达到预期值。这种缓慢采用的原因之一是驾驶员对范围感到焦虑,特别是对个人拥有的车辆而言。这里,用户经验差,例如常规汽车堵塞充电站或不连贯的实时可用数据,显示了驾驶员的焦虑范围。在这个背景下,我们研究可轻易地在当今导航系统中部署的随机搜索算法,以便尽可能减少绕行到一个可用的充电站。我们将这种搜索作为固定视野马尔科夫决策过程的模式,并提出一个考虑各种问题变异、加速技术和三种解决方案算法的综合框架:精确的标签算法、超常标签算法和推出算法。广泛的数字研究表明,我们的算法大大缩短了寻找免费充电站的预期时间,同时提高解决办法的质量,并且比近视方法成功的可能性。