Putting People in Their Places

5 Dec, 2007 By: Kenneth Wong

A STEPS animation sequence can help architects determine how crowds might behave in certain spaces.

There’s one aspect of AEC that has less to do with measurements and aesthetics and more to do with behavioral science. To outsiders, and even to some industry insiders, this specialized discipline seems more speculative than deductive, more artificial (as in artificial intelligence, or AI) than architectural. It’s computer-aided crowd movement simulation, often used to study how occupants might interact with the living space around them.

Game developers and filmmakers routinely use the same technology to digitally populate their scenes. But recent advances in hardware horsepower and visualization algorithms have pushed this technology toward a new science, now trusted by AEC consultancies. At SIGGRAPH 2007 in August, Massive Software, known for its AI engine in the video game and film industries, began courting the AEC market (see “ Animation Possibilities for AEC,” AEC Tech News 208).

But Mott MacDonald Group, an engineering consultancy, preceded Massive in this endeavor. Mott MacDonald developed Simulation of Transient Evacuation and Pedestrian Movements (STEPS), a software product described as “a microsimulation tool” to help its clients better understand pedestrian movements inside architectural structures.

According to the company, the technology “originates from the extensive experience of Mott MacDonald in the design of transportation systems, in particular, underground rail stations and interchanges, combined with its long experience in developing computer simulation tools for engineering design.”

Historical STEPS
“About 10 or 11 years ago,” recalled Norman Rhodes, vice-president of Hatch Mott MacDonald (Mott MacDonald’s North American arm), “we developed a fairly simplistic version of STEPS, mainly for evacuation processes.”

The software was intended to validate whether the design space complied with prevalent standards, such as those of the National Fire Protection Association (NFPA), which sets guidelines for how quickly the occupants in a building should be able to exit in case of emergency.

STEPS was originally developed to study evacuation scenarios. In this STEPS simulation, the second train car is on fire and smoke is propagating into the platform and station building through the platform side doors and windows on both sides of the car. People are evacuating via four sets of stairs (two on each platform).

This in-house project eventually grew to become a full-fledged product, now called STEPS. Like Massive’s AI engine, STEPS employs a collection of agents -- 3D digital human prototypes programmed to interact with nearby solid objects in a realistic fashion.

According to Mott MacDonald, the software's key features include

  • simulation of both normal and emergency operations
  • direct import of 2D and 3D CAD models
  • 3D interactive (virtual reality) graphical user interface
  • route system as alternative to the use of a cumbersome origin-destination matrix
  • the use of moving vehicles (such as trains and lifts)
  • pedestrian movement metrics with graphical representation

Studying Agent Smith
Rhodes explained the programmable agent’s behaviors in an evacuation as follows: “Each person, or agent, has a potential field of vision [a range of geometry it can detect], which guides it or compels it towards one exit or another."

For additional realism, the agent’s behavior can be fine-tuned. Baljinder Bassi, Hatch Mott MacDonald’s senior project manager, explained, “Here in New York City’s Pennsylvania Station, for example, in the morning you have lots of commuters passing through the station who are very familiar with their surroundings. They’re also impatient [as they’re trying to get to work]. So they have faster than normal walking speed. By contrast, in the afternoon, you have tourists who are not familiar with their surroundings. So they move slower. Families, on the other hand, tend to stick together. So in the simulation, those groups will move together. There’s also fatigue after traveling a long distance. You can program these types of attitudes and attributes by giving the agents a level of familiarity to govern their speed and movement.”

For computing the walking speed of a typical pedestrian, STEPS often uses NFPA’s standards as the starting point (Rhodes revealed that the standards call for a speed of one meter per second), but data on certain specialized segments of society are hard to come by. “There isn’t a lot of information on the average walking speed of people in wheelchairs, vision-impaired people, disabled children, or even children, for that matter,” Bassi said.

Mott MacDonald is currently partnering with the psychology department of the University of Pennsylvania to further develop STEPS.

The screen shows simulation of Penn Station during normal operation, with people entering and exiting the station via several routes. The experiment was conducted to ascertain the Fruin Level of Service Contours (a pedestrian placement concept advanced by John J. Fruin) at the concourse level.

Working with CAD
STEPS has its own built-in environment modeling tools, or 3D sketching tools, but they’re not as robust as those you might find in a CAD system. 3D Modeling is not STEPS’s purpose, Bassi pointed out. “So very early on, we gave STEPS the ability to bring in flat 2D drawings as DXF files. The original DXF format was flat; however, DXFs are no longer flat. Their lines and polylines can have elevations.”

If you have a 3D file, it can be imported into STEPS as an ASE file (used by Autodesk 3ds Max). “The advantage of the ASE format is that if the artist has created lots of detailed textures, like a marble floor, you can see them as such in the STEPS simulation, because the software has a good built-in rendering engine,” Bassi explained.

If the 3D model contains decorative objects, such as furniture, then these, too, can function as blockages or interferences that affect the behavior and movement of agents.

Animated Insight
If a picture is worth a thousand words, perhaps an animation sequence is worth many times more. The simulation results in STEPS could easily be summarized in a tabulated set of numbers, but the real power of the software, Rhodes pointed out, is in its ability to depict how people might behave.

“Quite early on, in conjunction with a building code consultant, we did a simple evacuation study for Edmonton International Airport in Canada,” Rhodes recalled. “When we ran the evacuation simulation using the standard walking speed [as defined by the local building code], when we looked at the numbers and the time it took for the people to exit the premises, we didn’t detect anything out of ordinary. But when we looked at the animation, we noticed everyone was queuing up at certain doorways. What that told us was that we could significantly improve the evacuation speed if we had put additional doors there, or widened the doorways.”

If you’re running computational fluid dynamics (CFD) analyses to study fire hazards, “you can even import the smoke from a CFD model,” Bassi explained. In other words, run a CFD session on the design space -- say, a retail store’s interior -- to see where smoke buildups would occur in case of fire. Then, with the smoke defined as interference, run an evacuation scenario in STEPS to see how the agents behave, which alternate paths they take to avoid the smoke, where the bottlenecks occur, and so on.

Rhodes explained the process as follows: “CFD data is imported into STEPS as a data file, giving the 3D concentration field, so that the impact on walking speed, for example, can be included in the simulation. 3D surfaces representing smoke concentration can also be imported and viewed as the STEPS simulation proceeds. Animations are made during the run.”

In addition to tabulated data and still images of the simulation in progress, STEPS can produce AVI or QuickTime movie files. If desired, the agents can be rendered with custom skin textures and accessories (baseball fans with logo-printed T-shirts or consumers with shopping carts, for example).

Licensed Agents
Hatch Mott MacDonald has done very little to promote STEPS as a commercial product. “We haven’t actively marketed it; we’re relying on word of mouth,” Rhodes said. He also observed that the clients who saw STEPS in operation often ended up licensing the software.

STEPS costs $6,000 per license annually. It includes the capability to model evacuation scenarios and normal pedestrian movements. An evacuation-only version is available for $4,500 per annum. (Note: This model has some limitations with regards to pedestrian modeling.) The annual license includes access to support and all upgrades.

About the Author: Kenneth Wong