How akshitshandilya/sih-26 works
This repository is a GNSS/INS outage-simulation and trajectory-analysis application. Its user workflow is to select a parsed drive and outage duration, hide GPS positions for that interval, predict the trajectory from inertial data, and inspect the predicted path and telemetry. Separate workflows prepare vehicle and smartphone logs, train inertial-feature models, and score predictions against withheld GPS ground truth. The graph reflects the sampled modules and documented subsystem roles; some wiring, particularly evaluation dispatch, is not fully visible in the excerpts.
Main entry points: processed Parsed drives · app.py Telemetry dashboard · harness.py Drift scoring · train.py Model training
Components 16
Data preparation
- Vehicle logs
data/raw - Phone logs
data/raw_s - Vehicle parser
src/pipeline/parse.py - Phone parser
src/pipeline/s_dataset.py - Parsed drives
data/processed
Trajectory prediction
- IMU features
src/pipeline/imu_features.py - Model training
src/models/train.py - Saved predictors
src/models/saved_models - Fusion filter
src/models/fusion_filter.py - EKF baseline
src/models/baseline_ekf.py
Evaluation
- Outage simulator
src/pipeline/outage_sim.py - Drift scoring
src/eval/harness.py - Score results
outputs/results.json
User application
- Telemetry dashboard
src/dashboard/app.py - Trajectory map
Other components
- Analyst
Connections 18
- Vehicle logs → Vehicle parser: reads logs
- Phone logs → Phone parser: reads logs
- Vehicle parser → Parsed drives: writes parsed drives
- Phone parser → Parsed drives: writes parsed drives
- Parsed drives → Model training: reads training data
- Model training → IMU features: builds features
- Model training → Saved predictors: writes predictors
- Parsed drives → Telemetry dashboard: loads selected drive
- Telemetry dashboard → Outage simulator: simulates outage
- Telemetry dashboard → Fusion filter: predicts trajectory
- Fusion filter → IMU features: builds features
- Fusion filter → Saved predictors: loads yaw model
- Telemetry dashboard → Trajectory map: renders paths
- Analyst → Telemetry dashboard: selects simulation
- Parsed drives → Drift scoring: loads evaluation drive
- Drift scoring → Outage simulator: simulates outage
- Drift scoring → Score results: saves scores
- Drift scoring → EKF baseline: uses predictor