A Scalable Agent Based Multi-modal Modeling Framework Using Real-Time Big-Data Sources for Cities

Abstract

This paper presents a framework for using real-time big-data to inform a transport Agent Based Model (ABM) for a range of scenario testing applications. Computational advances have enabled for increasingly complex, bottom-up, fine resolution simulations to be carried out over long time horizons at fine spatial and temporal resolution. This has hinted at the possibility of connecting scales of what has been historically been fine resolution operational models and coarse resolution strategic models. The value of any fine resolution dynamic model is limited by the quality of its inputs. The wave of new geospatially connected devices has enabled the harvesting of fine resolution spatial and temporal data on travellers’ and even the infrastructure itself. This crowd-sourced data can be used to inform dynamic models with real-world and real-time data, bypassing the need for generalised functions and/or expensive survey data. In this paper, Google Directions API data and Transport for London data feeds are presented in a framework for London. The use of decentralised data structures is also presented and comment is made on the possibilities of using parallel computing advances in Computer Science to scaling up fine resolution scenario testing transportation models and enabling support for a range of agent decision making methodologies. Such data structures offer performance improvements in the storing of dynamic data that may be manipulated in order to simulate local and global hard infrastructure scenarios alone or in tandem with traditional policy or dynamic policy making scenarios.

Gerry Casey

Principal Research Fellow, UCL & Associate, Arup

Transport modeller and data scientist building city-scale simulations to help governments and cities make better decisions on transport, climate, and equity.