A realistic software package gives designers and engineers a virtual vision of how building occupants would fare in a fire – allowing for amendments that could save many lives.
If the ghost of Sir Christopher Wren lingers in his visionary twin-towered hospital – now the University of Greenwich – it would do well to call in on Professor Ed Galea, director of the university's fire safety engineering group and developer of software BuildingEXODUS. The architect would have found this crowd simulation and evacuation software very useful in 1666, when he drafted his plan to rebuild London after the Great Fire.

But even the shade of Sir Christopher might be spooked by the images created by the virtual reality post-processor, which complements the latest edition of "the world's most advanced crowd simulation software".

Strange, plastic-faced manikins drift through fire-struck spaces. Many of these apparent zombies congregate around exit points. Some, showing no trace of pain, decelerate as they succumb to toxic fire gases. Others, chillingly, dissolve into the floor – signifying death. These sprites are a new development – on the main software, there are simply dots to represent human beings.

The program is based on tough engineering realities. And since its launch in 1996, the EXODUS model has provided building designers with a sophisticated means of testing evacuation procedures. Safety engineers, too, can now test more designs in less time to reach the optimal solution. The package's creators also stress that it can help to demonstrate that a building complies with performance-based standards.

Central to any evacuation plan is the layout of a building. So the EXODUS model works by allowing designers to import a standard CAD file of their floorplan into the software. This plan is then flooded by the dots representing people, and a virtual fire is started.

Each dot has a "personality" provided by the program. Obviously, they do not laugh, cry, or recite sonnets. But their attributes do include age, sex, speed, dexterity and patience.

What's more, in version 3.0 the dots can be credited with a localised knowledge of the building they inhabit. Most office dwellers, for example, are aware of the location of exits, but occupiers of public buildings are usually not.

EXODUS comprises five core sub-models: occupant, movement, behaviour, toxicity and hazard. A set of rules define the functions of each sub-model, which simulate the interaction between people and people, people and fire and people and structures.

The new model addresses smoke-filled environments, with dots exhibiting inefficient movement caused by reduced visibility. The poor dots stagger like drunkards, clinging to the geometry of the walls. Even more disturbing is the dots' ability to change colour from green to a red depending on their circumstances and how hot they are.

And, in the latest version, human behaviour has been modelled to show how, even in a dire emergency, occupants will pause to collect a jacket or handbag before attempting to leave. To some observers, this may seem bizarre. Surely people go into a frenzy of panic in a fire situation? Faced with Dante's inferno or the fate of Joan of Arc, who would care about their jacket? According to Professor Galea, quite a few.

And he hates the word "panic". Says Galea: "It's such an overused and overrated term. People are more ordered, more rational than one might think." He cites the Ladbroke Grove train disaster, on which he was chief fire consultant, as an example. "Even in the chaos and the hell of the crash, evidence suggested that human behaviour in crisis is far more positive and supportive than it's generally given credit for." He insists that the addition of the "itinerary" function, allowing individuals to perform simple tasks prior to exiting, extends the range of EXODUS "beyond emergency scenarios to normal circulation patterns in buildings". This, he believes, will help to predict how buildings and their users will interact in normal circumstances such as rush-hour in stations, and everyday crowds in shopping malls and sports stadia.

The behavioural information that inspired Galea and his team to invent EXODUS was not gathered overnight. It comes, in fact, from 15 years of research into people's reactions to fires, through accident reports and studies of actual fires. Survivors were interviewed.

Post-mortems were studied. And high on the wall of Galea's orderly office is a reminder of the software's ultimate purpose – a vivid photograph of the Bradford football stadium fire.

The software's most publicised application was at the Millennium Dome in Greenwich. When the virtual fire was started, the dome's building engineers were anxious to observe bottlenecking as potential escapees tried to get out. It was clear from the playback that a blockage was occurring around a point between exhibits. Alterations were made to the geometry of the escape route and the simulation revealed that escape times had been considerably reduced.

EXODUS has already been used by engineering consultants, architects and fire brigades in 17 countries around the world. Illustrious buildings on which it has been applied include World Expo in Portugal, the Sydney Olympic Stadium in Australia and Denmark's Esbjerg concert hall.

Galea is reluctant to go into his clients' design failings. But he did reveal that the biggest problems in building design are limited exit capacity, high travel distances and layouts that are too complex.

The emergence of version 3.0 is by no means the final chapter in the story of EXODUS – which, incidentally, can run on a standard desktop PC. Galea hints that a prototype model is being developed that simulates ever more complex group behaviour. For example, it incorporates features that show how status and hierarchies in offices produce natural leaders – shepherding their flocks to safety – and how family groups behave in emergencies.

That is in the future, but let us return briefly to Sir Christopher and the most celebrated fire in British history. Two-thirds of London's buildings succumbed in a "horrid bloody flame", but fewer than a dozen people died. Surely co-operation could be the only reason for this – historic proof, if that were needed, of Galea's theory about group behaviour in an emergency.