MATHEMATICS
tracksim
A reproducible sensor-fusion and multi-target tracking sandbox that exposes estimator behavior through a real-time HUD.
- Current state
- Simulation
- Discipline
- Mathematics
- Built with
- TypeScript / Vite / Canvas / Web Workers / Vitest / Playwright
The idea
Make estimator behavior visible under noise, missing measurements, clutter and association uncertainty in a reproducible synthetic environment.
How it works
Five simulated navigation sensors feed an extended Kalman filter with chi-square innovation gating. Simulated radar adds noisy detections and clutter; nearest-neighbour association and M-of-N confirmation maintain tracks. A fixed-step seeded model runs separately from the canvas presentation.
What’s implemented
- Extended Kalman fusion with Joseph-form covariance updates
- Innovation gating, covariance-derived influence and comparison against simulated ground truth
- Sensor degradation, fix-spoofing, clutter and decoy scenarios
- Seeded runs, backward scrubbing, headless execution and telemetry encoders
- Canvas HUD with a worker bridge and documented mathematical contracts
PROJECT STATUS / SIMULATION
Where it stands
Deterministic simulation sandbox
- All sensors and radar detections are simulated
- Simplified local 2-D kinematics and sensor models, not field-validated navigation or radar hardware
- Reproducibility describes the simulation implementation rather than real-world predictive accuracy
Source & resources
Documentation behind this project page
Reviewed October 1, 2026. Project descriptions reflect a source review; repository validation claims were not independently reproduced.