III : Small : Integrating and Learning on Spatial Data via Multi-Agent Simulation

人类在城市或地区的流动对经济活动、社会互动和基础设施发展至关重要。 虽然流动的某些方面可以而且已经在微观一级进行了研究，而且可以收集区域统计特性，但提供一个符合微观动态的全球规模整体模型的任务仍然是一项挑战。 为了建立一个完整的经济和社会后果模型，必须收集社会人口数据，并跟踪各种运输方式和时间范围内的流动情况。 优化电动汽车充电基础设施和促进公平的交通执法等直接影响人们日常生活的紧迫挑战将受益于这些数据和易于使用的模型。 为了实现这些目标，该项目将建立一个人类流动的基础模型。该模型可以包含各种类型的运动，时间和社会人口统计数据。 它还可以生成与微观尺度事件或宏观尺度属性相关的输出，并且与两者一致。 该模型的核心是一个模拟引擎，它可以以与所有用作输入的数据一致的方式模拟事件和运动。 这种表示的学习将利用现代机器学习技术，同时优化以与不同类型的数据模态保持一致。由此产生的模拟模型利用时空异常检测的创新来研究干预和预测任务。最终，这些有形和可访问的模型将有助于通过计算、数据科学和土木工程之间的新联合项目培养下一代空间科学家，从而促进STEM的广泛参与。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The movement of humans throughout a city or region is critical to economic activity, social interactions, and infrastructure development. While some aspects of mobility can and have been studied at the microscopic level, and regional statistical properties could be gathered, the task of providing a holistic model for a global scale that adheres consistently with micro-scale dynamics has remained a challenge. For a full model of the economic and societal consequences, it is essential to gather socio-demographic data and to track movements across various modes of transportation and time scales. Pressing challenges, such as optimizing electric vehicle charging infrastructure and promoting equitable transportation law enforcement, that directly impact people’s daily lives, will benefit from access to such data and easily-usable models. Towards these goals, this project will build a foundational model for human mobility. This model can incorporate diverse types of movement, temporal, and socio-demographic data. It can also generate outputs relevant to a micro-scale event, or a macro-scale property, and consistent with both. At the core of this model is a simulation engine that can imitate events and movements in a way that is consistent with all of the data used as input. The learning of this representation will utilize modern machine learning techniques, and simultaneously optimize to align with different types of data modalities. The resulting simulation model leverages innovations in spatio-temporal anomaly detection to study intervention and prediction tasks. Ultimately these tangible and accessible models will help facilitate broadening participation in STEM by developing the next generation of spatial scientists through new joint programs between computing, data science, and civil engineering.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.