SC Telemetry Tester

Simulates the onboard Raspberry Pi CAN reader and replays CSV/Excel logs into InfluxDB for deployment testing. This replaces the former data_replayer/ Python tool in race-profile.

Architecture

Production:   Firmware → Pi (can-telem-cloud) → InfluxDB → inference API
This tester:  CSV / Excel / generated data → InfluxDB → inference API

Setup

npm install

Configure InfluxDB (same env vars as docker/main.py):

export INFLUXDB_URL="https://us-east-1-1.aws.cloud2.influxdata.com"
export INFLUXDB_TOKEN="your-token"
export INFLUXDB_ORG="your-org"
export INFLUXDB_BUCKET="your-bucket"

See .env.example for optional settings.

Write modes

Mode	Env	Line protocol
production (default)	INFLUX_WRITE_MODE=production	telemetry_snapshot,signal=soc value=85.5 <ts_ns> — matches the car
flat	INFLUX_WRITE_MODE=flat	telemetry,source=sc_telemetry_tester soc=85.5,pack_power=1500 <ts_ns> — matches current docker /predict defaults

For end-to-end inference tests against the deployment microservice, use flat mode until /predict queries the production signal tag layout.

Replay modes

Mode	Command	Behavior
Bulk (default)	npm run batch	Writes all rows using CSV timestamps (Var1 ms epoch if present)
Bulk + recent window	--shift-to-now	Rewrites timestamps to start at now; use with /predict lookback_seconds
Real-time	npm start	Sleeps --interval seconds between rows; loops by default

Examples

From race-profile root (after git submodule update --init):

cd tools/sc-telemetry-tester
npm install

# Bulk load training CSV, shifted to current time (works with lookback_seconds)
export INFLUX_WRITE_MODE=flat
node src/replay.js \
  --file ../../data_pipeline/dataProcess/testData/raw_data.csv \
  --shift-to-now

# Simulate live 1 Hz telemetry from generated test data
npm start

# Custom fields (flat mode)
node src/replay.js \
  --file data/test_telemetry.csv \
  --fields soc,pack_power,air_temp,speed \
  --realtime --interval 1 --loop

# Excel replay
node src/replay.js --file path/to/log.xlsx --shift-to-now

CLI flags

Flag	Description
--file	CSV or Excel path (default: data/test_telemetry.csv)
--fields	Comma-separated fields for flat mode
--measurement	Override measurement name
--timestamp-column	Timestamp column (auto-detects Var1, timestamp, time, _time, Time (s))
--interval	Seconds between rows in realtime mode (default: 1)
--realtime	Sleep between rows
--loop	Restart from first row after file ends
--shift-to-now	Rewrite CSV timestamps relative to now
--max-rows	Cap rows written
--start-row	Skip leading rows (e.g. pre-race idle telemetry with speed 0)
--route-file	Grades/route CSV (latitude, longitude, elevation_m, distance_m) — injects lat/lon/elev into each telemetry row
--route-mode	distance (speed × Δt along route, default) or index (row fraction along route)
--speed-field	Telemetry column for speed in distance mode (default: speed)
--speed-unit	mph (default) or mps

Route + signal fusion (old telemetry, 2026 course GPS)

When legacy logs have missing GPS (lat/lon = 0 or -1001) but good CAN signals, replay signals from TestDataAnalysis and inject location from BSR ASC grades CSV:

export INFLUX_WRITE_MODE=production
export INFLUXDB_MEASUREMENT=telemetry_snapshot

node src/replay.js \
  --file ../../course_data/TestDataAnalysis/2024-4-7rawdata.csv \
  --route-file "../../course_data/BSR ASC GPX Grades Data 1.0 2026/UofMtoSummitElmentary_grades.csv" \
  --route-mode distance \
  --speed-unit mph \
  --start-row 31267 \
  --shift-to-now \
  --max-rows 6000

2024-4-7rawdata.csv has ~31k idle rows at speed 0 before driving starts — use --start-row 31267 so distance mode advances along the route.

--max-rows 6000 at ~20 Hz ≈ 5 minutes — enough for one strategy lookback window (300 s).

Full segment (~83k rows) without --max-rows is fine for bulk load; strategy uses the latest 300 s window.

Use --route-mode index for a quick test without integrating speed.

End-to-end test (race-profile)

1. Load telemetry (terminal 1):

   cd tools/sc-telemetry-tester
   export INFLUX_WRITE_MODE=flat
   node src/replay.js \
     --file ../../data_pipeline/dataProcess/testData/raw_data.csv \
     --shift-to-now

2. Start inference (terminal 2): see docker/README.md

3. Query (terminal 3):

   curl -X POST http://127.0.0.1:8000/predict \
     -H "Content-Type: application/json" \
     -d '{"lookback_seconds": 300, "fields": ["soc", "pack_power", "air_temp"]}'

Other commands

npm run generate        # Create test_telemetry.csv from format.json
npm test                # Verify Influx connection and write permission

File structure

sc-telemetry-tester/
├── data/
│   ├── data_format.json      # Fallback copy of signal schema
│   └── test_telemetry.csv    # Generated test drive cycle
├── src/
│   ├── influx_client.js      # Influx v2 line-protocol writer
│   ├── load_replay_file.js   # CSV / Excel loader
│   ├── timestamp.js          # Timestamp detection and shift-to-now
│   ├── replay.js             # Main replay CLI
│   ├── route_inject.js       # Merge route grades CSV → lat/lon/elev
│   ├── generate_test_data.js
│   └── test_connection.js
└── package.json

Data format

Signal definitions live in sc-data-format (format.json). When cloned as a sibling under tools/ in race-profile:

tools/sc-data-format/format.json   ← canonical
tools/sc-telemetry-tester/          ← this repo

Do not nest sc-data-format as a submodule inside this repo; keep them as flat siblings in race-profile.