menorCaminho

Work in progress. This is a study project, not a finished product. The Chinese routing algorithm (Tsinghua) is not implemented yet; city mode, live replan, and behavioral noise are still pending.

Central goal: build realistic traffic scenarios and measure how well routing algorithms optimize travel under congestion, with the main target being maximum efficiency of the Chinese (Tsinghua) algorithm compared to baselines (BFS, Dijkstra) on the same map, demand, and agent count.

Python simulator: urban network with lanes, signals, and FIFO queues. Plug-and-play GPS algorithms run in a test matrix; reports include fairness metrics (spread, Gini, p95) and comparative victory signals.

Not a product. A lab to stress-test routing under load before and after the Tsinghua integration.

Current stage

Done	Pending
Realistic district map (realistic-1.1)	Tsinghua GPS (main objective)
placeholder (BFS) and dijkstra (assign-only)	Live replan from telemetry
Event-driven GpsService, test matrix, fairness reports	City-scale map mode
Parallel session runs (max 4 workers)	Behavioral noise, OSM import

What it does today

• Generates a synthetic urban district (hierarchical roads, flow devices)
• Simulates N vehicles with per-lane FIFO queues and effort in seconds
• Compares plug-and-play GPS algorithms under the same stress matrix (seed, demand, N)
• Reports fairness metrics and victory signals vs the placeholder baseline

Run

python -m venv .venv
.\.venv\Scripts\Activate.ps1
python src/main.py

Details: docs/getting-started.md

Available GPS (today)

Name	Strategy
placeholder	BFS, assign-only (baseline)
dijkstra	Shortest path by static effort (devices + t_free, no live queue)

Layout

src/
  traffic_code/   laws and parameters
  graph/          map, effort, traversal
  simulation/     controller, demand, reporting
  algorithms/     GpsService + implementations
docs/             architecture, tests, criteria

License

MIT: free to use; personal/academic study, no warranty.