We study a new variant of the vehicle routing problem, called the Mobile Production Vehicle Routing Problem (MoP-VRP). In this problem, vehicles are equipped with 3D printers, and production takes place on the way to the customer. The objective is to minimize the weighted cost incurred by travel and delay of service. We formulate a Mixed Integer Programming (MIP) model and develop an Adaptive Large Neighbourhood Search (ALNS) heuristic for this problem. To show the advantage of mobile production, we compare the problem with the Central Production Vehicle Routing Problem (CP-VRP), where production takes place in a central depot. We also propose an efficient ALNS for the CP-VRP. We generate benchmark instances based on Vehicle Routing Problem with Time Windows (VRPTW) benchmark instances, and realistic instances based on real-life data provided by the Danish Company 3D Printhuset. Overall, the proposed ALNS for both problems are efficient, and we solve instances up to 200 customers within a short computational time. We test different scenarios with varying numbers of machines in each vehicle, as well as different production time. The results show that these are the key factors that influence travel and delay costs. The key advantage of mobile production is flexibility: it can shorten the time span from the start of production to the delivery of products, and at the same time lower delivery costs. Moreover, long-term cost estimations show that this technology has low operation costs and thus is feasible in real life practice.
Learning from Drivers to Tackle the Amazon Last Mile Routing Research Challenge
Amazon Web Services（亚马逊云科技）
Amazon Last Mile Routing Research Challenge的目标是将Amazon司机的真实体验整合到最佳路线规划和优化的解决方案中。本文提出了我们的方法，通过分层结合机器学习和传统的旅行商问题（TSP）求解器来应对这一挑战。我们的方法从这两个世界中获益。一方面，我们的方法通过从区域级别的历史路线中学习顺序概率模型来编码驾驶员的专业知识，其中每个区域包含一些包裹停靠站。然后，它使用一个单步的策略迭代方法，称为推出算法，以生成合理的区域序列从学习的概率模型采样。另一方面，我们的方法利用丰富的TSP文献中开发的成熟方法，有效地在每个区域内进行排序。这种结合的结果似乎是有希望的。我们的方法获得了$0.0374$的评估分数，这与官方挑战排行榜上的前三名球队所取得的成绩相当。此外，我们基于学习的方法适用于可能表现出超出本挑战范围的不同顺序模式的驾驶路线。我们的方法的源代码可在https://github.com/aws-samples/amazon-sagemaker-amazon-routing-challenge-sol上公开获得。
An Optimization Framework for the Time-Dependent Electric Vehicle Routing Problem with Shared Mobility: A Step Toward Smart Cities
Cornell University（康奈尔大学）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；The University of Sydney（悉尼大学）；Sharif University of Technology（谢里夫理工大学）
The Combined Van-and-Mopeds Routing Problem with Time Windows: Formulation, Variants, and Solution Methods
Instituto Superior Técnico, Universidade de Lisboa（里斯本大学高等技术学院）；INESC-ID Lisboa（里斯本嵌入式与系统创新研究中心）；Dingoo（丁果）；Instituto Superior de Engenharia de Lisboa, Instituto Politécnico de Lisboa（里斯本理工学院高等工程学院）
