The Social Force Model: How Crowds Self-Organise and Jam

Treating people like particles with a destination

The Social Force Model, introduced by Dirk Helbing and Péter Molnár in 1995, describes each pedestrian as a point particle acted on by a small set of forces, and simulates crowd motion by integrating Newton's second law for all of them at once. Unlike a real particle, the driving force does not come from a field — it comes from the pedestrian's own intent, constantly steering them toward a chosen exit or goal at their preferred walking speed. Three terms make up the total force on person i:

F_i = F_drive + Σ F_social(i,j) + Σ F_wall(i,k)

F_drive        = (v0_i · ê_goal - v_i) / τ         pulls toward desired speed & direction
F_social(i,j)  = A · exp[(r_ij - d_ij) / B] · n̂_ij  repels from other pedestrians j
F_wall(i,k)    = same exponential form, repels from walls/obstacles k
  (+ body-compression and sliding-friction terms once agents physically touch)

The social force F_social is an exponentially decaying repulsion — strong at close range, fading quickly with distance — that keeps pedestrians from walking through each other and encodes personal space. It is deliberately not a hard collision constraint: like flocking or traffic models, the group-level structure (lanes, gaps, avoidance) emerges from many simple, local, particle-level interactions rather than being scripted in.

Lane formation and the emergent order

Run the model with two streams of pedestrians walking in opposite directions through a corridor and, without anyone deciding to do it, the crowd self-organises into alternating lanes of same-direction traffic — the social-force repulsion makes it cheaper, moment to moment, to fall in behind someone going your way than to keep colliding with oncoming walkers. The same self-organisation produces oscillating flow at bottleneck doors (traffic alternates in bursts through a narrow gap) and stripe patterns at diagonal crossings. None of these patterns exist in the equations explicitly; they are collective outcomes of the pairwise repulsion.

Panic and the faster-is-slower effect

The model's best-known and most counter-intuitive result appears in evacuations. Raise everyone's desired speed v0 (simulating panic) and, up to a point, the exit flow rate increases as expected. Push it further and the flow rate drops — pedestrians converging on a single exit compress against each other hard enough to form a self-sustaining, roughly semi-circular arch of jammed bodies around the doorway, which physically blocks the opening faster than people can pass through it. This is the faster-is-slower effect: individually rational urgency produces a collectively worse outcome, matching the clogging and stampede dynamics observed in real crowd-crush disasters.

Holding it back with design, not willpower

Because the arching is a geometric, force-driven phenomenon and not a psychological one, the fixes that work in the model are physical: widening the exit reduces arch pressure per unit width faster than linearly; placing an obstacle — a pillar or barrier — slightly in front of and offset from a doorway breaks up the arch before it forms, counter-intuitively increasing flow by reducing the number of people who can simultaneously compete for the opening; and multiple smaller exits distributed around a room outperform a single large one, because they cap how many people converge on any one bottleneck at a time. These are now standard features of building-code egress design, validated in part through Social Force Model simulations of exactly the kind used here.

Frequently asked questions

Try it live

Everything above runs in your browser — open Pedestrian Flow and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.