Notes

Analyses, simulation studies, and commentary from the pedestrian dynamics community.

November 18, 2026 · 13 min read

Calibrating a pedestrian model against real data, part 3: default values for JuPedSim 2.0

JuPedSim 2.0 renames its models and needs default parameters that a user can start from. We recalibrate the Collision Free Speed model on two experiments at once, add a vision factor against clogging, and test the result on data it was not fitted to. The proposed set does not clog in any of our tests. However, it underestimates the flow in wide bottlenecks by about 20 %.

Read more →

November 4, 2026 · 8 min read

A column at the top of a stair

At Copenhagen Airport a column stands right where people turn onto a stair. I rebuilt the spot in JuPedSim. In the simulations, the column pushes people towards the glass, makes them step onto the stair at an angle, and slows a crowd down.

Read more →

October 21, 2026 · 24 min read

When can we trust a crowd simulation? Lessons from a road paradox

A good explanation makes specific consequences follow from explicit assumptions; evidence decides where those consequences can be trusted. A road paradox shows the first part, a calibration study of a crowd model the second, along with what simulation standards do and do not establish.

Read more →

October 10, 2026 · 6 min read

Can a model move like museum visitors and still take the wrong paths?

From a phone video of the Guggenheim Bilbao atrium we extracted 331 visitor trajectories and rebuilt the scene in JuPedSim. The model matches how visitors move, but not where they spend their time.

Read more →

September 7, 2026 · 20 min read

Calibrating a pedestrian model against real data, part 2: where it stops

The parameters calibrated on a wide bottleneck in part 1 go to a half-metre gate and an unguided entrance. One set stalls, the other is too fast, a recalibration disagrees with the first, and no tested static parameter slice reproduces low motivation. What that means for anyone who runs pedestrian simulations.

Read more →

September 5, 2026 · 30 min read

Calibrating a pedestrian model against real data, part 1: the bottleneck

We calibrated JuPedSim’s Collision Free Speed model with Dakota on an open bottleneck experiment. The defaults miss the flow by a factor of two; calibration brings it mostly within ten percent, but two optimizer starts give two different parameter sets.

Read more →

June 29, 2026 · 8 min read

Why do roaming crowds drift counterclockwise, and where?

A Nature Communications paper reports that freely roaming crowds drift counterclockwise. We rebuilt the experiment in simulation to test whether a slight individual left-turn bias explains it.

Read more →

June 12, 2026 · 4 min read

Why don't airlines board planes the optimal way?

The mathematically optimal boarding method has existed since 2008. We rebuilt the classic studies in JuPedSim to see why it stays on paper.

Read more →