EAGER: Collaborative Research: Towards the Development of Smart Bike Sharing Systems

移动和基于传感器的技术的最新进展使收集和处理各种人类流动性数据成为可能。当与环境信息和交通信息相结合时，这种人类流动性数据可以用于开发具有更广泛社会影响的重要应用程序。该项目通过专注于两个数据分析挑战，考虑了在城市内构建高效的自行车共享系统所出现的技术问题。一是单车在不同站点的需求量预测。二是不同站点间的最优单车再平衡策略。自行车需求的成功预测可以帮助系统运营商更好地部署自行车，并在车站之间重新分配自行车。有效和优化的自行车再平衡可以帮助满足自行车租赁的动态需求，节省系统运行成本。尽管主要关注自行车共享网络，但通过此问题改进的数据分析功能应该适用于其他类型的分布式租赁服务的问题。这一探索性研究项目旨在开发有效的、可扩展的数据挖掘和优化技术，具有预测不同站点的自行车需求和优化站点间自行车再平衡策略的分析能力。首先，该项目旨在开发基于回归的预测模型，该模型同时考虑相关特征和上下文信息，如共享单车站点之间的连接。其次，研究了以最小化车辆总行程为目标的自行车再平衡问题的混合整数非线性规划(MINLP)方法。虽然传统的MINLP技术不能保证可行的解决方案，但研究小组的目标是开发先进的集群技术，首先将站点分组为集群，然后使用集群来促进MINLP。该项目还为评估所开发解决方案的有效性制定了适当的措施。该项目为研究生和本科生提供基于研究的高级培训机会。该项目产生的所有数据、软件和出版物将向更广泛的研究社区公开。

Recent advances in mobile and sensor-based techniques have made it possible to collect and process a variety of human mobility data. When combined with environment information and transportation information, such human mobility data can be used to develop important applications with broader societal impacts. This project considers technical problems that arise in the context of building an efficient system for bike sharing within a city by concentrating on two data analytics challenges. The first is the prediction of the demand of bikes at different stations. The second is the optimal bike rebalancing strategy among different stations. The successful prediction of bike demand could help system operators better deploy bikes and redistribute bikes among stations. Effective and optimal bike rebalancing could help meet the dynamic need of bike rental and save system operational costs. Although primarily focusing on bike sharing networks, data analytics capability advanced through this problem should be applicable to problems from other types of distributed rental services. This exploratory research project aims to develop effective and scalable data mining and optimization techniques that have the analytical capability to predict bike demand of different stations and to optimize the bike rebalancing strategy among stations. First, this project aims to develop regression-based prediction models that take into account both relevant features and contextual information such as connections among bike sharing stations. Second, this project explores mixed integer nonlinear programming (MINLP) techniques for solving bike rebalancing problem with the objective of minimizing the total travel distance of rebalancing vehicle. While traditional MINLP techniques could not guarantee feasible solutions, the research team aims to develop advanced clustering techniques to first group stations into clusters and then use the clusters to facilitate MINLP. This project also develops appropriate measures for assessing the effectiveness of the developed solutions. The project offers research based advanced training opportunities for graduate and undergraduate students. All the data, software, and publications resulting from the project will be made publicly available to the broader research community.