At present I have/am contributing to following research projects at CASA:

	Developing a Prototype Agent-Based Pedestrian Evacuation Model with Repast
	The Use of New 3D Visualization Software in Reassurance Policing
	Geodemographics for Managing Local Services

Agent-Based Modelling of Pedestrian Evacuation: A Study of London's King's Cross Underground Station

London’s King’s Cross St. Pancras underground station has been the unfortunate location of two major incidents within the last twenty years. A fire in November 1987 and the terrorist bombings in July 2005 both resulted in the loss of lives, and the injury of many people. The implementation of measures to mitigate or neutralise the effect of any future incidents at this site is unrealistic. The adoption of preparedness measures is crucial for the emergency services to limit the loss of life and property, and to improve the response phase of an incident. The King’s Cross underground station is currently being redeveloped, partly to mitigate the remaining few operational and safety issues raised after the 1987 fire, and also to allow for the future increases in passenger volume (e.g. 2012 Olympics). However, despite these modifications and improvements, both the surrounding built environment and the station will necessarily remain complex structures. The emergency services have several duties placed upon themselves in the event of a major incident, and a computer based simulation tool capable of examining the effects of different incident assumptions or contingencies may greatly benefit an incident planner.

The aim of this research is to design and implement a prototype pedestrian evacuation model in order to facilitate the assessment of local National Health Service (NHS) resources in the event of an incident within or adjacent to King’s Cross Underground station. In particular, Camden PCT are interested in the appraisal of pedestrian egress from the station in order to determine the allocation and positioning of key emergency functions and facilities e.g. ambulance loading point(s), casualty clearing station(s) to which the injured can be taken, etc.

Computer simulation offers an efficient means of modelling the interaction of a large number of autonomous entities, especially when the evaluation of different contingencies is required. At present, there are no fewer than thirty three proprietary and non-proprietary computer simulation applications / models available to evaluate pedestrian egress from buildings. An agent-based pedestrian evacuation model has been developed for this study using the Repast (Java) toolkit, which was identified as a viable alternative to using an existing pedestrian evacuation application / model for several reasons. Off-the-shelf proprietary pedestrian evacuation models often provide limited explanation of their inner workings. Many are essentially black box, and accompanying literature often provides little or no evidence to support the validity of the results they produce. Additionally, accessibility to some proprietary pedestrian evacuation models is limited to consultancy with the software developer. In such instances, the party who has commissioned the research invariably has limited understanding of the modelling techniques used, and are limited to the results and analysis published in the final report. They are unable to explore results further, or revisit a model of a scenario for additional exploration at a later date. Neither of these options is particularly desirable to the research sponsor.

Project Website

Innovative use of the BCS 2000 allows surface representation and analysis of fear of crime at the local level. BCS data has been coded at the unit postcode level (i.e. ~ 15 households) and extrapolated across the study area using the MOSAIC UK geodemographic typology to provide a significant insight into the spatial distribution of fear of crime. MOSAIC UK is a neighbourhood classification built from statistical cluster output of a wide range of demographic, socio-economic and geographic data also at the unit postcode resolution.

The natural surveillance of the study area was calculated from a 3D model constructed from an intelligent digital map supplied by Ordnance Survey (OS) of Great Britain for use with Geographical Information Systems (GIS) and databases. MasterMap includes topographic information on every landscape feature i.e. buildings, roads, phone boxes, postboxes, landmarks. Combined with Light Detection and Ranging (LiDAR) data, which provides very high accuracy elevation data acquired at approximately 1m spatial resolution with a height accuracy of around 10 to 20cm, a very detailed profile of the built environment can be constructed quickly, and cost effectively. By separating buildings from other feature classes within the MasterMap data, their approximate height can be calculated from the average elevation of observation points contained by each footprint. Subsequently, individual point features (e.g. trees, street lights, CCTV) can be represented by predefined 3D symbols within the software package (figure 1).

View-shed analysis was used to calculate the visibility of every location within the 3D model, thus evaluating the natural surveillance of the area. Results from the model clearly identify narrow alleyways with limited natural surveillance in red, whilst large visually unobstructed areas are depicted as green. Figure 2 shows a raster grid beneath the 3D environment displaying the fear of crime index for London. Areas identified with both a poor natural surveillance and increased fears of crime were subsequently reviewed for the presence of street lighting and CCTV. Further view-shed analysis of the street lights and CCTV coverage identified hot-spot areas (locations with a decreased likelihood of been observable), which could benefit from increased crime prevention tactics.

	Figure 1 (top left): Line of sight analysis of CCTV cameras and street light coverage. Figure 2 (top right): Fear of crime index displayed beneath the 3D model of London’s King’s Cross. Figure 3 (bottom left): Fully rendered 3D landscape.

he visualization of the 3D model can be developed further by incorporating techniques, such as photo modeling, tin buffering (using draped aerial photography), or further predefined 3D symbols (i.e. buildings) to include more detailed geometries (figure 3).

This work builds upon knowledge and expertise gained during the development of a demonstrator produced with the Environmental Systems Research Institute (ESRI) Inc., for their 2004 user conference plenary. Public sector collaborations include the British Transport Police, Metropolitan Police Service (London Borough of Camden), the London borough of Camden Council, and Camden Primary Care Trust.

Geodemographic profiles of individuals, households and small areas are already recognised to be pivotal to tactical and strategic resource management in many areas of business, and could become similarly central to efficient and effective deployment of resources by public services. Profiles might be created using new and existing sources of public sector data alongside detailed, locationally disaggregate and frequently updated data from commerce. Such sources remain severely under-used in academic and public service research, as recognised in the March 2003 Commission on the Social Sciences report. Data sources that are created within many of the public services could be used to provide valuable context to decision-making, but remain neglected by potential users within and around public services.

This research sets out an agenda for adapting best practices for developing and using geodemographics in public service delivery, and identifies how a geographic information system (GIS) can crystallise the motivations and impediments to such transfer. A specific empirical objective of this research is to establish a demonstrator GIS that, though pooling of public and private sector data, could serve the needs of diverse interest groups that have a shared interest in the regeneration of the King’s Cross site – one of the largest urban redevelopment projects in the EU this decade – whilst minimising disruption and harm to the area’s existing communities. Developments in the case study site include construction of the Channel Tunnel link and terminus, diversion of local road networks and utilities, and substantial retail and residential redevelopment.

To achieve this objective an integrated georeferenced data set for Camden / Islington area of North London has been built, with the intention of making it available through an interactive project Website. It is expected that the Website will include a 3D visualisation of the study area, and will make it possible for stakeholders to address issues such as street crime, adaptation of retail structure and the health impact analysis of the development. Central to accomplishing this is the pooling of data from a range of partners in the public and private sectors.

The research suggests that, hitherto, stark differences in working environments and practices have masked the commonalties of interest in geodemographic analysis extant within and between the public and private sectors. An objective of this study is to create a research network that crystallises these commonalties amongst local stakeholders, and thus makes possible an improved evidence base for local public service delivery. Viewed prospectively, this research also suggests a catalytic role for the academic sector in defining and achieving important strategic synergies across the public-private divide.

Project Website

Return to Top