Modelling techniques

(updated  31st August 2007)

Keywords, Pedestrian Evacuation, simulating people, simulating crowds, simulating crowd dynamics, workshop

We have been running model building workshops for the last six years. It never ceased to amaze me that people will apply the wrong techniques to a problem simply because they have invested in the wrong tools and therefore need to use them. As the saying goes :

"When all you have is a hammer - everything looks like a nail."

At all times a model should adhere to the BATNEEC principle.

Myriad is a multi-scalar methodology. A technique where the appropriate modelling tools are used in the appropriate situations. We use queueing models, where appropriate, flow models, where appropriate and many other techniques as the requirement demands. This approach relies more on common sense that heavy investment in a specific tool.

The VEgAS system modelled individuals in 3D. Simulex in 2D, Steps in 3D and our own validation tools  to model uni- and bi-directional flows. Once you have built a wide range of models of complex geometry it becomes obvious the many microscopic analysis (individual elements) can be replaced with the appropriate macroscale formulae. For example, there is no need to build models of turnstiles at a microscopic level as there exists a vast wealth of data for the mean and standard deviation of the flow rates though turnstiles. This is HOW the building codes and safety regulations are evolved and applied. Through observation, experimentation and careful analysis.

Apply a formula for that problem, perhaps coupled to a visibility analysis for approach and a queueing model for processing (arrival v service rates) and you can solve most transport terminal problems. This advantage of using this type of methodology is that it takes minutes - not hours - to build, test, analyses and understand the models.

Setting up a microscopic model for the above can take many hours/days and would require several modelling runs to provide an answer you may already know (or can find out by simply asking).

Over the last six years many very good, easy to use and extensively tested models have been built, tested, validated and documented. For example, below we have a very small selection of the systems we have used over the years.

"Crowd Dynamics" Validation tool (PhD thesis models above left) illustrating high density, bi-directional flow in a corridor

Above right MiS module from the Myriad II suite - Jamarat Bridge.

VEgAS models (above right and below right) of high density movement through complex geometries.

One can apply microscopic modelling for such problems as high density, bi-directional flow (see below) and you can also apply macroscopic modelling.

There are a number of papers published on these techniques and the reader is recommended to see Pedestrian and Evacuation Dynamics, ISBN 3-540-42690-6

Microscopic modelling has its place as a research tool and it may be useful in some "unknown" geometries but that has to be weighed up against the cost of modelling and the validity of the outputs. Over the last 12 years we have researched many tools and developed an extensive model building experience for high density, bi-directional crowd environments such as stadia, railways, shopping malls, exhibitions, religious gatherings and many more. Details of this technique are available on request and outline of the research click here for further information.

Extract from Dr. Still PhD thesis Crowd Dynamics

Myriad combines elements of both micro- and macroscopic modelling in an integrated package. It reads data files produced by third party microscopic models and can be used to generate easy to understand graphic for reports.

Whether your requirements are moving dots on the screen, full blown Virtual Reality, evacuation analysis, spatial design and analysis, crowd dynamics, pedestrian planning and design tools, services, consultancy etc. you should ask the following questions of any vendor - including ourselves (of course).

1.  What validation do you offer ?

2.  Is this a black-box or an open source model ?

3.  How long does it take to build a model ?

4.  How can we test/validate the underlying assumptions in the model ?

5.  How brittle is the model - if I make a small change how long does it take to change the model ?

6.  What is the cost of building and modifying a model ?

If you are not satisfied with the answers - caveat emptor. At all times a model should adhere to the BATNEEC principle.