gui4gmns

(in the *4gmns family alongside plot4gmns / path4gmns / osm2gmns)

AI-guided dashboards for GMNS: DTALite · TAPLite · Dynamic ODME · DLSim. GMNS run folder in → one self-contained, offline-capable dashboard.html out. No install to view, no editing to learn — the legacy NEXTA GUI stays the editor; this is the modern viewer + generator.

gui4gmns is the project name — the pip-installable generator core plus its viewers. It is the cross-platform successor of the Windows-only MFC NEXTA GUI, rebuilt on one shared data contract with an AI-guided generator at its core. See NAMING.md.

The process — decompose it forward, transfer to any city

Five steps forward — osm2gmns → demand → assignment → integrate → visualize — then swap the city's inputs and the steps + dashboards stay the same. Full how-to: docs/GMNS_TO_DASHBOARD_SKILL.md.

Quickstart

# 1) install (pure-Python core, no required deps)
pip install gui4gmns
gui4gmns datasets/01_sioux_falls                           # -> datasets/01_sioux_falls/dashboard.html
#    options: --basemap osm|satellite|none   --max-traj N   --single

# 2) or import it, like plot4gmns
python -c "from gui4gmns import generate; generate('datasets/01_sioux_falls')"

# no install? run the generator straight from the repo instead:
python ai-gen/gui4gmns.py datasets/01_sioux_falls

# 3) open the dashboard (double-click, or serve for the basemap)
python -m http.server 8765   # then browse to datasets/01_sioux_falls/dashboard.html

What a generated dashboard gives you

• Network map with MOE coloring/bandwidth: volume · V/C · queue · time-dependent flow · QVDF speed
• Hybrid basemap: OSM network-wide, satellite as you zoom into detail — both embedded (offline)
• Animation: play/scrub 15-min bins; vehicle dots (green moving / red queued) from trajectories
• Demand OD desire lines + demand matrix heatmap and attribute distributions (learned from plot4gmns — see ai-gen/LEARNINGS_FROM_PLOT4GMNS.md)
• Corridor speed contour: INRIX-observed vs QVDF-model space-time, with RMSE / R² / bias validation
• Data-quality audit in every dashboard (MOE coverage, conservation, oversaturation, suspicious zero-volume major links, connector sentinels) — it checks its own inputs
• Split lightweight layer files (dashboard_layers/*.js) so each layer is inspectable & regenerable

Repository layout (this is the canonical home — release + continuous updates happen here)

ai-gen/                the generator core (gui4gmns.py) + VIZ_SCHEMA + design studies
nexta_x.html           web-lite viewer (zero-install, drag-drop)
web-gl/                GPU viewer (regional-scale animation, live-follow)
desktop-qt/            Qt desktop app (Run engine button, basemaps, headless snapshots)
engine/DLSim_STE/      the DLSim space-time-event engine (C++17 source + toy testdata; build.sh)
engine/bin/            local binaries (git-ignored; build from source or copy dlsim_run.exe here)
datasets/              public samples: 01 Sioux Falls · 02 Chicago Sketch · 05 toys · 07 West Jordan
docs/                  Users Guide (md + pdf)
SHARED_CONTRACT.md     the one data contract every branch implements

Provenance: consolidated 2026-07 from the dtalite/ (gui4gmns, DLSim_STE), the DTALite/TAPLite C++ kernel workspace, and dynamic_ODME/ sample sets, so releases happen in ONE place. Heavy kernels stay in their dev homes: DTALite/TAPLite C++ → dtalite_with_taplite_Cpp_kernel/ · large networks → asu-trans-ai-lab/dynamic-odme-lab.

The viewers (one shared contract, SHARED_CONTRACT.md)

Visualization portals — online + offline

GMNS is a shared exchange format — think DICOM in medical imaging: you don't read the raw file, you open it in a portal. gui4gmns opens one GMNS folder in a whole family of them. Offline (no internet): the dashboard, web-lite, web-gl, desktop-qt, QGIS, Google Earth Pro, static figures. Online (richer 3D + trajectories): Kepler.gl, deck.gl, Google Earth. One command exports outward to all four:

python exporters/gmns_to_viz.py <gmns_folder> --target all    # -> kepler + deckgl + qgis + kml, each with a README

Every generated dashboard auto-writes a portals/ folder beside it (kepler/deckgl/qgis/kml) — one run, every portal (--no-portals to skip). Full guide — what each portal is for, 3D & trajectories, and how students swap in their own datasets: docs/VISUALIZATION_PORTALS.md.

▶ Live demo (no install — opens in your browser): a deck.gl view of the Chicago Sketch network at https://asu-trans-ai-lab.github.io/gui4gmns/portal_demo/ · browse the gallery at https://asu-trans-ai-lab.github.io/gui4gmns/gallery.html.

Data policy

This repository ships the generator, viewers, and public sample networks only. Agency data (NVTA / VDOT / INRIX / CBI and any QVDF corridor derived from them) is never committed — it is git-ignored via *PRIVATE* / *nvta* rules. Large public networks (ARC Atlanta, Chicago Regional) are referenced from asu-trans-ai-lab/dynamic-odme-lab rather than duplicated here. Run python validate_no_private_data.py before any commit.

Sample-data attribution: Sioux Falls / Chicago from bstabler/TransportationNetworks; basemap tiles © OpenStreetMap contributors and Imagery © Esri, Maxar, Earthstar Geographics (embedded per their terms). The local-only ITS I-95 (VA) data-hub demo is sourced from the USDOT JPO CodeHub Data Cleaning and Fusion Tool (https://github.com/usdot-jpo-codehub/data-cleaning-and-fusion-tool); its INRIX/VDOT/probe layers are restricted and never committed.

Docs

docs/VISUALIZATION_PORTALS.md (online+offline portals, 3D/trajectories, student how-to) · docs/gui4gmns_Users_Guide.md · SHARED_CONTRACT.md · REFACTOR_PLAN.md · ai-gen/VIZ_SCHEMA.md (build your own dashboard, AI-guided) · ai-gen/LEARNINGS_FROM_PLOT4GMNS.md.