Roundstone Crossing Predictor 🚂

Predicts when the barriers at Roundstone Level Crossing (Angmering, West Sussex) will be open or closed, using live Network Rail train data.

How it works

1. Listens to Network Rail's real-time STOMP feeds (Train Describer + TRUST)
2. Tracks individual trains approaching the crossing from both directions
3. Infers crossing barrier state (open/closing/closed/opening) with confidence levels
4. Logs every state change to SQLite for historical analysis

Crossing Details

• Location: Roundstone Level Crossing, B2140, East Preston/Angmering
• Type: MCB-CCTV (Manually Controlled Barriers with CCTV)
• Railway: West Coastway Line (BLI1), 70 mph
• Between: Angmering (ANG) ↔ Goring-by-Sea (GBS)
• Traffic: ~176 trains/day

Setup

Prerequisites

• Python 3.11+
• Network Rail Open Data account (register here)

Install

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env with your NROD credentials

Run

python -m src.main          # Start the predictor
python -m src.main --api    # Start with API server

Web Dashboard

Open http://localhost:8590 in your browser for a live dashboard with:

• Schematic track diagram with train positions
• Predictions tab with upcoming closure windows
• Upcoming trains panel (from Realtime Trains)
• Crossing state history
• Feedback form in the footer

Docker Deployment

# On your server:
./update.sh              # Pull image and start container
docker compose logs -f   # View logs

The app is deployed via GitHub Actions — every push to main builds a Docker image and pushes to GHCR.

API

GET /                      — Web dashboard
GET /status                — Current crossing state + confidence
GET /diagram               — Tracked trains with berth positions (for schematic)
GET /predictions           — Upcoming trains + predicted closure windows
GET /predictions/windows   — Upcoming closure windows from RTT data
GET /next?station=         — Upcoming services at a station (ANG or GBS)
GET /history               — Query historical open/close intervals
GET /stats                 — Summary statistics
GET /health                — System health (uptime, feed status, DB size)
GET /up                    — Simple uptime check (200 + JSON uptime)
GET /sf/summary            — S-Class signalling summary
GET /sf                    — S-Class signalling events
POST /feedback             — Submit feedback (JSON: {"message": "..."})
GET /feedback              — Retrieve feedback (requires Authorization: Bearer <token>)

Architecture

NROD STOMP ──→ Train Tracker ──→ Crossing Inferrer ──→ API + Logger
  (TD+TRUST)    (per-train)      (state + confidence)   (FastAPI + SQLite)

States

State	Meaning
UNKNOWN	No data / just started
OPEN	No trains approaching, crossing clear
CLOSING_PREDICTED	Train detected in approach zone, closure expected in ~Xs
CLOSED_INFERRED	Train at/near crossing, barriers likely down
OPENING_PREDICTED	Train passed, barriers likely rising
STALE_DATA	Feed connection lost, state unreliable

Licence

MIT