TASHA/2: Next Generation Agent-Based Microsimulation

基于活动/旅游的城市出行需求模型越来越多地被用于业务规划实践。这些通常是在基于代理的微观模拟(ABM)框架内实施的，在该框架中，为单个出行者(代理)建立了户外活动参与和相关旅行的模型。ABM提供了一种极其灵活、强大和有效的方法来模拟复杂的时空和社会经济行为，如旅行。这类模式的一个主要例子是Tasha，自2016年以来一直在大多伦多地区(GTA)投入运营。 关于各种来源的出行行为的“大数据”的迅速增长、计算能力的持续增长以及不断变化(且日益具有挑战性)的政策分析问题(自动驾驶车辆、新的移动服务、促进积极的交通运输、社会公平问题等)。创造机会，也有必要继续开发更先进、更稳健的旅行需求建模方法，以帮助引导世界各地城市区域沿着更可持续的道路继续爆炸性增长。 目前的型号代表着“第一代”ABM。然而，在新出现的大数据集(以及其他信息源)和高性能计算的基础上开发更强大的“第二代”ABM存在相当大的需求和机会。然而，要做到这一点，将涉及开发新的行为表示和计算算法，在更加灵活的软件环境中实施，既充分利用可用的计算能力，又能够灵活地试验和扩展新的交通方式和服务以及不断演变的出行行为的表示。 拟议研究的目标是开发这样一个第二代反弹道导弹模型系统：Tasha/2。在Tasha已证实的基地设计的基础上，下一个版本将解决上述确定的许多弱点。这项研究将围绕6个相互关联的主题展开： 改进了旅行决策的代表性。 利用新的数据源，结合数据融合和机器学习方法，以新的方式建立新的模型。 开发一系列移动服务的显式模型，这些服务在代理人的旅行选择和通过物理运输网络模拟车辆和人员的移动之间进行“调解”。 “机动性工具”选择的明确建模(拥有汽车、传统和自动驾驶、汽车共享会员、共享单车会员等)。 明确表示停车位的供应和选择。 使用先进的计算方法，确保TASHA/2能够在实际可接受的运行时间内模拟复杂的运输过程。

Activity/tour-based models of urban travel demand are increasingly used in operational planning practice. These are generally implemented within an agent-based microsimulation (ABM) framework, in which out-of-home activity participation and associated travel are modelled for individual trip-makers (agents). ABM provides an extremely flexible, powerful and efficient means for modelling complex spatial-temporal, socio-economic behaviour such as travel. A leading example of such models is TASHA, which has been in operational use in the Greater Toronto Area (GTA) since 2016. The rapidly growing availability of “big data” concerning travel behaviour from a variety of sources, continuing growth in computing capabilities, and ever-changing (and increasingly challenging) policy analysis issues (autonomous vehicles, new mobility services, promoting active transportation, social equity concerns, etc.) create both the opportunity and the need to continue to develop more advanced, robust travel demand modelling methods to help guide the continuing explosive growth of urban regions worldwide along more sustainable paths. Current models represent “first-generation” ABMs. Considerable need and opportunity, however, exists for the development of significantly more powerful “second-generation” ABMs that build upon emerging big datasets (among other information sources) and High Performance Computing. To do so, however, will involve the development of new behavioural representations and computational algorithms, implemented within much more flexible software environments that both fully exploit available computing power and enable flexible experimentation with and extension of representations of new transportation modes and services and evolving travel behaviour. The objective of the proposed research is to develop such a second-generation ABM model system: TASHA/2. Building upon the proven base design of TASHA, the next version will address many of the weaknesses identified above. The research will be organized within 6 interconnecting themes: Improved representation of travel decision-making. Exploiting new data sources in combination with data fusion and machine learning methods to build new models in new ways. Developing explicit modelling of a range of mobility services which “mediate” between agents' travel choices and simulation of the movement of vehicles and people through the physical transportation network. Explicit modelling of “mobility tool” choices (car ownership conventional and autonomous, car-sharing membership, bike-sharing membership, etc.). Explicit representation of parking supply and choice. Use of advanced computing methods to ensure that TASHA/2 can model complex transportation processes within practically acceptable run times.