NEXt generation activity and travel behavioUr modelS: Bringing together choice modelling, ubiquitous computing and data science

在世界上许多国家，运输部门占公共支出的很大一部分。例如，2018年英国在交通方面的公共支出总额为225亿英镑。新的交通决策的潜在影响可以使用数学模型来评估，以预测人们将做什么，何时何地，以及在任何给定场景下他们将如何在不同地点之间旅行。这些旅行行为模型通常基于经济学和心理学理论，并利用调查数据开发。然而，新的出行方式（如自动驾驶汽车、优步、共享单车）和新型用户（如老年旅行者、移民）正在导致出行格局发生根本性变化。传统的数据和模型无法处理日益复杂的活动和旅行模式，而这正是NEXUS的动力所在。当前主流模型的局限性是多方面的。首先，他们假设旅行行为完全取决于旅行者的年龄、收入、态度等，以及其他选择的属性（如旅行时间、成本）。他们没有考虑到可能影响个人决定的无数心理因素，例如压力、疲劳或更普遍的“思考过程”的影响。其次，用于开发模型的数据通常依赖于小规模调查，在这些调查中，旅行者被要求报告/记录他们过去的行为，或者根据假设场景的描述陈述他们的选择，这些通常不是真实世界旅行行为的可靠衡量标准。与此同时，大量的移动数据不断从GPS、手机和社交媒体等来源产生。先进的技术和机器学习（ML）方法也使得通过简单的腕带、离散的夹子和基于智能手机的传感器来测量旅行者的“精神状态”成为可能，并从大脑成像中推断他们的思维过程。此外，虚拟现实（VR）技术的进步使旅行者能够沉浸在未来的场景中，以获得更真实的反应。将新的数据和方法结合在一起可以导致旅行行为建模的一个步骤变化——但是统一这些不同研究流的框架尚未形成。NEXUS建议通过开发方法，用新颖的数据形式来增强旅行行为模型，以解决这一研究缺口。这将包括：(a)从GPS、移动电话和其他被动来源产生的真实移动数据；(b)使用传感器测量的有关“精神状态”的动态数据；(c)未来虚拟现实场景下的旅行行为实验数据。利用被动移动数据和感知心理状态将涉及使用最先进的机器学习和无处不在的计算技术。结合不同类型的真实世界和实验数据源来预测新场景中的行为，将涉及将这些数据整合到传统的旅行行为建模框架中。首次将这些技术结合到实验室之外，将产生一套更丰富的出行行为模型，可以更好地应对未来截然不同的交通场景和用户群体。这些模型将在微观模拟平台中实施，以提高准确性模拟不同政策情景下的流动性行为，并帮助规划者和政策制定者做出更明智的投资决策。这项多学科研究将建立并扩展我过去使用大数据和传感器进行行为建模的经验。它将支持我过渡到利兹大学的研究领导角色，并与交通、心理学和计算机领域的全球知名学者合作。与非学术合作伙伴的伙伴关系将确保研究快速过渡到实践和现实世界的影响。

In many countries around the world, the transport sector claims a major share of the public spending. For example, the total public spending on transport in the UK was £22.5 billion in 2018. The potential impacts of new transport decisions can be evaluated using mathematical models to predict what people will do, when and where, and how they will travel in-between different locations in any given scenario. These travel behaviour models are typically based on theories of economics and psychology and developed using survey data. However, new forms of mobility (e.g. self-driving cars, Uber, shared-bikes) and new types of users (e.g. older travellers, migrants) are leading to radical changes in the mobility landscape. The traditional data and models are failing to deal with the rising complexities of activity and travel patterns which motivates NEXUS. The limitations of the current mainstream models arise from multiple factors. Firstly, they assume travel behaviour is solely based on the age, income, attitudes, etc. of the traveller and the attributes of the alternatives (e.g. travel times, costs). They do not account for the myriad of psychological factors that could influence an individual's decision, for example, the effect of stress, fatigue or the 'thinking process' more generally. Secondly, the data used for developing the models typically rely on small-scale surveys where travellers are asked to report/log their past behaviour or to state their choices based on descriptions of hypothetical scenarios, which very often are not reliable measures of the real-world travel behaviour. On a parallel stream, large amounts of mobility data are constantly generated from sources like GPS, mobile phones and social media. Advanced technologies and machine learning (ML) methods have also made it possible to measure the 'mental state' of the travellers by simple wristbands, discrete clip-ons and smartphone-based sensors and infer their thinking processes from brain imaging. Further, advances in virtual reality (VR) technology has made it possible to immerse travellers in future scenarios to obtain more realistic responses. Bringing together new data and methodologies can lead to a step change in travel behaviour modelling - but the framework to unify these different streams of research is yet to be formulated. NEXUS proposes to address this research gap by developing methodologies to augment travel behaviour models with novel forms of data. These will include: (a) real-world mobility data generated from GPS, mobile phones and other passive sources; (b) dynamic data about the 'state-of-the-mind' measured using sensors; and (c) experimental data on travel behaviour from VR settings of hypothetical future scenarios. Utilizing passive mobility data and sensing mental states will involve utilizing state-of-the-art ML and ubiquitous computing techniques. Combining the different types of real-world and experimental data sources for predicting behaviour in new scenarios will involve integrating these in traditional travel behaviour modelling framework. Merging these techniques, for the very first time outside the lab-setting, will produce a richer set of travel behaviour models that can better deal with radically different transport scenarios and user-groups in the future. The models will be implemented in a microsimulation platform to simulate the mobility behaviour in different policy scenarios with increased accuracy and aid the planners and policy-makers in making more informed investment decisions. This multi-disciplinary research will build on and extend my past experience in behavioural modelling using big data and sensors. It will support my transition to a research leadership role at the University of Leeds and collaboration with globally renowned academics in transport, psychology and computing. Partnership with non-academic partners will ensure the quick transition of the research to practice and real-world impact.