The MATSim loop

In one sentence

The converged state is what scenarios are compared in; calibration tunes the parameters so that it matches observation.

Why it matters

Every person keeps up to five plans. Each iteration one plan is executed on the network and timetable, scored, and in replanning either kept (85 %) or changed — a new route, shifted times, another mode for a subtour (5 % each). Over 400 iterations the plan memory fills, the worst plan is forgotten, and the average score flattens; for the last 20 % innovation is switched off.

Agent-based simulation is usually explained with equations or with a screenshot of a running model. Neither shows why a planner should care. These explainers show the mechanism itself — one person, one day, one score — so that the results on a dashboard stop being a black box.

Sources

• Scoring parameters are typical MATSim values (illustrative); see Horni, Nagel & Axhausen (2016), ch. 3
• Horni, A., Nagel, K. & Axhausen, K. W. (eds.) (2016): The Multi-Agent Transport Simulation MATSim. Ubiquity Press

What you see

1. 00.0Plans → simulation → scoring → replanning
2. 02.0Plan memory fills
3. 06.0Average score rises
4. 14.0Innovation off, last 20 %
5. 17.0Converged

Use this explainer

MP4 (1800×720) · GIF (1200 px) · Poster (PNG)

Embed: <iframe src="https://www.replan.city/explains/embed/matsim-loop.html?lang=en" width="900" height="360" style="border:0;max-width:100%;aspect-ratio:5/2" loading="lazy" title="The MATSim loop"></iframe> Copy
Paste into any page; the explainer plays as live SVG (no video file).

Licence: CC BY 4.0 — reuse freely with the line “Replan GmbH · replan.city/explains”. The Replan mark in the corner stays.