Like shuffling along in a crowd? No? Simulation to the rescue!

Jan 28, 2019 | Hot Topics

At last year’s Bentley Year in Infrastructure, I was introduced to something I had never seen before: pedestrian modeling. In my naïveté, I assumed that engineers observed humans moving through spaces and used that to inform how they design subway stations, hotel lobbies, shopping malls and other places where people congregate. They do — but then they turn that into mathematical models to enable prediction and optimization.

At YII 2018, Bentley announced that it had acquired LEGION, whose pedestrian simulation application models people’s interactions with one other and with physical obstacles in places like railway stations, airports, sports arenas and at street level, with vehicles.

It gets even better. While I was talking with LEGION’s founder, Douglas Connor, he and his colleague showed me how they simulate people getting on airplanes. I fly a lot and can attest; whether from the windows towards the aisle, back to front, or (as sometimes happens) utterly at random, boarding is a mess. Until we all stop trying to squeeze overfilled bags into the overhead lockers, it’s not going to be easier. But with tools like LEGION*, the designers of aircraft interiors can at least try to make it better.

Mr. Connor told me that how people move, purposefully or randomly, singly and in groups, is a key part of their perception of their surroundings. He believes that these flows are fundamental design criteria for public spaces. And, he said, using LEGION’s simulations with Bentley’s building and civil design modeling solutions means pedestrians will be considered “from the strategic and capital planning stages of a project, throughout its design, retrofits, and into asset operations.”

When I tweeted about LEGION from YII, a number of you responded with names of other pedestrian modeling providers– and, it turns out, there are quite a few.

And it makes sense. If you’ve ever been at a user conference when 5,000 people are trying to leave an amphitheater at once, you know the problems being simulated: will that pillar impede safe evacuation? Is that trash bin in the way? How do we manage the crowd so that everyone keeps moving at a pace supportable by all? (Even those looking down at their phones — don’t DO that, people. Walk or surf, not both.)

If you’re a retailer, you want to know where in the city center to locate a shop that draws the most walk-in customers. Town planners need to model flows through a park to correctly place sidewalks. How can an amusement park get people to the attractions in the back — or will they all stop at the rides at the entrance?

If you can model people flows, you can model other types of flows. MATSim, for example, is an open-source framework for transport simulations. A group in Santiago, Chile created a MATSim scenario that combined road data from OpenStreetMap, public transport data via Google Transit Feed Specification (GTFS) and travel diaries from a city-wide origin-destination survey to study how cars and buses traveled through the city. They were able to model how population changes would affect the city’s roads and residents, whether staggering working hours would be enough to deal with bottlenecks, and what impact tolls might have.

In Germany, there’s SUMO, or Simulation of Urban MObility, an open source multi-modal traffic simulation engine. Where MATSim is intended for an entire network, SUMO moves a single vehicle through a given road network. Each vehicle is modeled explicitly, with its own route. Why? To figure out traffic light patterns, evaluate traffic surveillance methods and forecast traffic loads.

The magic happens when you add all of these together: how pedestrians enter a crosswalk, how cars enter those same intersections, and design technology that can keep both safe. One possible solution is PORCA, the Pedestrian Optimal Reciprocal Collision Avoidance model, developed by the National University of Singapore. PORCA predicts a pedestrian’s “global navigation intention” –how they intend to get to a specific location– as well as any interactions the pedestrian might have with vehicles or other pedestrians. Autonomous vehicles have lots of sensors to detect objects around them, but can’t know which pedestrian is likely to dart out into traffic to get to the shop across the street. PORCA is a start at identifying pedestrians’ intentions to create a planning algorithm for the autonomous vehicle. The paper linked above says that, as of its summer 2018 publication, a robot scooter was able to drive safely, through a crowd.

I hadn’t heard of pedestrian modeling before YII 2018 and, now that I’ve had time to investigate the concept a bit more, Bentley’s acquisition of LEGION seems even more important. It’s important for urban planners and architects to understand how their spaces will be used, and to create a pedestrian experience that matches their intention. If people can’t get through the train station, or a theater can’t be safely evacuated, the design has failed. Crowd simulation should be a key part of every project.

It goes further than that, though, with the advent of autonomous vehicles. We typically talk about the car and its sensors, and how many billions of miles of driving are required to prove the technology. The infrastructure components –accurate maps, sensors on infrastructure that speak to the car, modeling how people share the roadways– are just as important. Projects like PORCA are starting to connect these dots, and I can’t wait to see what happens next.

Image is by Riccardo Bresciani, sourced from pexels.com