Transport Modelling — District Traffic What-Ifs In 3D

August 30, 2026

Ask "what happens if we close Oxford Street for an hour?" or "cut a lane on Commercial Road and drop the speed limit" and Alfrada OS runs a paired microscopic traffic scenario on an OpenStreetMap extract — baseline versus your interventions — then opens a pitched 3D replay in the Dashboards panel with vehicle meshes, congestion colouring, and KPI deltas. This is an experimental Vissim-style microscopic view powered by Eclipse SUMO, not a calibrated city forecast.

What you can do

• Centre + radius studies (≤ 5 km). Give a lat/lon and radius; Alfrada OS pulls a cached OSM extract, builds a passenger road network with street names, and runs SUMO.
• Paired baseline vs scenario. One reproducible synthetic demand set (seeded) drives both runs so deltas are fair — including unfinished vehicles.
• Interventions that stick to exact OSM street names. Full closures, lane reductions, and speed limits. Misspellings return close matches instead of silently guessing the wrong road.
• Demand knobs without rebuilding the network. Scale scenario demand with demand_multiplier and tune vehicles/hour/lane-km for the shared route set.
• A 3D replay dashboard. MapLibre + deck.gl pitched scene with OpenFreeMap buildings, play/pause/scrub/speed, baseline↔scenario switch, road hover, and travel-time / time-loss / waiting / speed / completion deltas.
• Three artifacts in the session. dash_transport_*.html (Dashboards), transport_*.json (KPIs + matched edges), and a ZIP with SUMO inputs/outputs plus OSM/SUMO attribution — all indexed for later search.

Where this shows up

• Event diversion planning. "Close Main Street for a parade and show me the delay impact in a 1.5 km radius."
• Lane-drop construction windows. "Reduce Commercial Road by one lane for 45 minutes and compare average speed."
• Safety speed trials. "Drop the limit on High Street to 20 km/h and show waiting time deltas."
• Stakeholder demos. Open the Dashboards tab and scrub the 3D replay instead of shipping a static map screenshot.
• Illustrative optioneering before a real Visum/Vissim study. Use Alfrada OS for directionally useful what-ifs, then hand the ZIP to a transport modeller for calibrated work.

Try it

• "Model traffic around Shoreditch High Street (51.525, -0.078) within 1.2 km. Close 'Shoreditch High Street' for 30 minutes and open the 3D replay."
• "Around 51.5074, -0.1278 with a 2 km radius, reduce 'Strand' by one lane and set demand_multiplier to 1.2."
• "Compare a 20 km/h speed limit on 'Borough High Street' versus baseline for a 20-minute run."
• "What if we close both 'Queen Victoria Street' and 'Cannon Street' — show KPI deltas and the dashboard."

Heads up

• Illustrative demand, not a forecast. Synthetic routes are reproducible and labelled as such — do not treat KPIs as calibrated AADT or scheme appraisal numbers.
• Exact OSM street names required. The tool suggests close matches on miss; it will not fuzzy-apply an intervention to the wrong road.
• District scale only. Max radius 5 km, horizon 20–120 minutes with a 10-minute departure cooldown so KPI means cover finished trips only. Larger studies belong in specialist software.
• Street-name discovery. action: list_roads returns the exact OSM names SUMO keeps for an area, with lanes, length and travel directions, and warms the network cache for the run.
• Time windows and one-way. Interventions take start_minute / end_minute (live rerouters and speed signs) and a one_way type keeps one travel direction open. Trips that would start or end on a closed street inside its window are dropped from both sides and reported as unserved.
• Raw results for charting. Per-vehicle and per-edge SUMO XML land in the session so the Code Executor can graph them.
• Beta / experimental. Needs the transport sidecar (SUMO 1.27.1). First runs over Overpass can take a few minutes; repeats reuse the OSM cache.
• Google Maps is not the network source. Place search stays on google_maps; road geometry for simulation comes from OpenStreetMap only.