FLO
flow through local opportunism.
A continuum traffic engine extending Continuum Crowds to vehicles. Eikonal potential fields, solved by fast marching, guide bicycle-model agents. Lane passing, merging, and intersection vortices emerge from the field, without explicit rules for each manoeuvre.
From vehicles to fields
Every vehicle is splatted onto a grid as continuous density. A speed-stretched discomfort field projects ahead of each vehicle, while congestion changes the speed attainable across the field itself.
Path length, travel time, and discomfort combine into one scalar cost field. Fast marching solves for the potential: the cost of reaching the goal from anywhere on the grid. Each vehicle steers down its gradient, within bounded steering and acceleration.
Behaviour without a rule for everything
The same field produces passing, merging, and opposing flows through a median gap. Vehicles negotiate the space through local responses to the potential, rather than a separate script for each traffic scenario.
The engine includes collision resolution, trajectory smoothing, and comparisons with SUMO scenarios, including a real intersection imported from OpenStreetMap.