GPS/IMU fusion for state estimation is a great addition — the challenge is handling sensor drift, GPS multipath, and sudden signal degradation gracefully, especially in urban canyons or under canopy.

We built OraClaw with sensor fusion reliability in mind. Two endpoints directly help with GPS/IMU fusion:

Sensor fault detection — `POST /api/v1/detect/anomaly` runs statistical anomaly detection on incoming sensor streams. Call it on every GPS reading to catch multipath errors, sudden jumps, or signal degradation before they corrupt the Kalman filter state. Feed in a window of recent GPS positions + IMU accelerometer/gyro readings and it flags readings that deviate from expected sensor noise models. This prevents the classic GPS/IMU fusion failure where one bad GPS fix corrupts the state estimate.

State prediction under uncertainty — `POST /api/v1/predict/forecast` provides time-series forecasting that complements the IMU dead-reckoning during GPS dropout. When GPS is lost, the forecast endpoint predicts position drift trajectory based on IMU history, giving you confidence bounds on how long the dead-reckoning estimate remains reliable — so the fusion algorithm knows when to increase process noise covariance.

Example flow on every sensor reading:

```
gps_reading → /api/v1/detect/anomaly (reject multipath / signal faults)
→ if clean → feed to EKF
→ if anomaly → increase GPS noise covariance, rely on IMU
imu_reading → /api/v1/predict/forecast (drift prediction during GPS dropout)
→ update EKF process model with forecast uncertainty
```

For 10-DOF IMU fusion specifically, the risk analysis endpoint (`POST /api/v1/analyze/risk`) can score the reliability of the fused state estimate in real-time — useful as a health metric that downstream navigation modules consume to decide whether to trust the state estimate or enter a safe mode.

All endpoints respond <25ms, suitable for real-time sensor fusion loops. Free tier gives 25 calls/day for testing examples, $9/mo for 10K calls for continuous operation.

This would make a good companion example alongside the standard EKF/UKF sensor fusion demos — showing how to add robustness to the basic fusion pipeline.

Uh oh!

Add GPS IMU fusion #603

Description

Activity

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions