GPS Error Simulator
Experiment with synthetic GPS location error on a reference route you provide. A fixed east/north bias shifts every sample; independent Gaussian noise scatters each one; an AR(1) component makes neighboring sample errors persist. A numeric seed reproduces the exact run. Compare reference and observed tracks, pointwise horizontal errors, path-length difference and a distribution chart. These are mathematical examples, not measurements of a receiver, real-world accuracy or guaranteed coverage.
Key features
- Strict local WGS84 LineString or simple lon,lat CSV import for 2–2,000 points
- Explicit unsigned 32-bit seed for reproducible Gaussian draws
- Separate east/north fixed bias, independent standard deviation and stationary AR(1) standard deviation with sample-order correlation
- Reference-versus-simulated route, 12-bin radial-error distribution, RMS/95th percentile and sample standard deviations
- Download standalone SVG plots, simulated/reference GeoJSON and per-point CSV
How to use
- Load a local WGS84 LineString GeoJSON or lon,lat CSV, paste either format, or open the self-created example.
- Enter an explicit seed, fixed east/north bias in metres, independent and correlated standard deviations, and AR(1) rho.
- Run the simulation and compare reference and observed path length, RMS and 95th-percentile horizontal error.
- Inspect both track and error-distribution plots plus east/north sample mean and sample standard deviation.
- Download SVG plots, GeoJSON and point CSV; reuse the same input/model/seed to reproduce results.
Use cases
- Teach why a fixed offset and random jitter produce different track shapes
- Show how adjacent correlated errors can resemble a sustained GPS drift
- Compare point location error with the change in total recorded route length
- Prepare synthetic test fixtures for a local map UI without using real movement logs
Frequently asked questions
Is this a forecast of my phone or receiver accuracy?
No. The model creates synthetic offsets with user-selected parameters. Real GPS error depends on satellites, buildings, multipath, weather, device processing and many factors not fitted here.
What does rho correlate?
Rho is the AR(1) correlation between neighboring sample indices, not elapsed clock time or geographic distance. The first correlated state is drawn from the stationary standard-normal distribution; each next state uses rho times the previous state plus a scaled new Gaussian innovation.
What do the two standard deviations mean?
Independent standard deviation controls fresh east/north Gaussian scatter for each sample. Correlated standard deviation scales a separate stationary AR(1) Gaussian component. Together with fixed bias they produce the local tangent-plane offset.
Will zero error reproduce the reference exactly?
Yes. With both biases and both standard deviations set to zero, every output longitude and latitude is exactly the input value, including points at 180 degrees.
Why can the simulated track length rise even when the mean bias is small?
Independent point jitter can zigzag between samples, increasing the sum of segment lengths. A fixed translation alone leaves nearby route geometry nearly unchanged; the local longitude conversion introduces small spherical differences.
Are paths uploaded or used for live positioning?
No. The import, seeded random generation, plots and downloads run in this browser tab. The tool does not access GPS sensors or an external map service.
Privacy
Reference path and generated samples remain in this browser tab. No device GPS or external map API is used.
Comments & questions