Station Evacuation - Analysis

(updated  31st August 2007)

Keywords: simulating people, simulating crowds, simulating crowd dynamics, workshops

Testing the evacuation of a busy London Underground station is a formidable task. Myriad was designed to assess the pedestrian impact for both normal and emergency situations and the diagrams below were produced from both the Myriad interface and Myriad running on a CAD plan of the site.

Myriad produces several different types of maps indicating congestion, hesitation points, ingress/egress, circulation and queueing. The model on the left shows the Myriad Level of Service map for a busy station (see diagram below left for the whole area showing the interaction zones (areas where multi-directional flow occurs).

The process of analysing a station concourse for emergency egress.

Step 1. CAD Plan to Isophotic map

This takes few minutes from a CAD DXF file (below left) imported into Myriad (below right) then click on the Process Image, flood fill and grey scales (each grey scale represents a different height level in the map.

Step 2. Run the analysis.

This section of the programme is semi-automated. Using a point and click interface it produces the following maps. We know the physical geometry from the dimensions and can either use demand data (O/D matrix) or assess the capacity of the area from the ingress/egress capacity (physical geometry).

Above - the congestion maps produces in a few minutes using the Myriad point and click interface. Click here for the Level of Service scale. This is the probability of conflict maps and indicates the overall spatial dynamics under normal operating conditions.

Step 3. Evacuation analysis.

Determining the local spatial dynamics we can then run a evacuation model and the superposition of both the spatial dynamics and the egress model highlight those areas we need to focus on for an emergency evacuation.

Rapid Results

Within 1 hour (from CAD to simulation results) we can produce Speed, Density and Space Utilisation simulation and animations for a station evacuation (see below)

Above - Speed map (colour represents speed of individuals) we can also produce Density map (colour represents density of individual)

Above - Space Utilisation map (colour represents the space used by people - see PhD thesis)

DWELL Analysis

Our agent analysis tools allow the user to zoom (dynamically) into the models. Clicking on the screen changes the display from agents to density, speed, space utilisation and position maps.

The agent models show density, speeds, space utilisation and agent location. The information is stored as a file so the data can be displayed in 3D CAD tools (such as 3D Studio Max).

Below - levels of service keyed to the agent locations - users can observe the dynamics of crowd density during the simulation

This modelling tool is provided with validation from field data.

Click here to download one page flyer

Throughput Analysis

The Myriad suite allow us to test for stability in flow down any route through the system under both normal and emergency situations. The graph below is indicative of a meta-stable system in which the crowd flow is periodic in nature and suffers from high clustering/interactions (ie: a turbulent journey through the system with a lot of stop-start flows). This is indicative of an unbalanced flow system (multiple cross flows, congestion impairing flow, stop/start movement of passengers in the concourse). By highlighting this kind of crowd dynamic we can address the problem during the design phase.

The Myriad suite is highlighting the inefficiencies in the design, namely that there are several "pinch-points" where passengers encounter cross flows and congestion. A more efficient system would sustain a higher peak, have a smoother transition curve and a steep rise and descent (looking like a plateau rather than a mountain ridge). Designing spatial dynamics is a science and this is rapidly gaining momentum with a wide range of multi-disciplinary engineers and consultants around the world.

These models took less than 1 hour to produce.