Mi Teleférico — Agent-based simulation
Statistically validated queueing model of the Red Line, with a 3D web twin
Simulation of passenger flow through the La Paz–El Alto cable car system using agent-based modelling. The model reproduces the three real stations of the Red Line as a queueing system with origin→destination routing, and was validated against twelve field measurement sessions. On top of it I built a 3D web simulation where each passenger picks a destination, boards a cabin and gets off at their station.
Highlights
- Validated with a Welch t-test over 30 runs: p = 0.932 against real-world data, so the null hypothesis holds and the model is considered reliable.
- BehaviorSpace experiments pinpointed the saturation point: opening a third ticket booth per station cuts total waiting time by 36% (~18s to ~11s); a fourth adds nothing.
- 3D web twin with React Three Fiber: cable, cabins and agents rendered in real time over the line's actual topology.
- Reproducible Python analysis (pandas + SciPy) that regenerates the validation from the field and simulation CSVs.
- NetLogo
- BehaviorSpace
- Python
- Next.js
- React Three Fiber
- Three.js