Walkthrough of building range-only relative localization for wearable devices using UWB ranging and IMU data fused via an Extended Kalman Filter.
Key Takeaways
Single-antenna UWB gives distance via time-of-flight but no bearing; PDoA adds bearing but requires two antennas, making wearable integration impractical.
External trilateration with fixed anchors (like GPS) solves location cleanly but requires infrastructure the system explicitly avoids.
EKF chosen because it handles nonlinear measurement functions (square root for distance) and works with one UWB antenna and two devices only.
State vector tracks delta-x and delta-y in world frame; IMU acceleration is control input, UWB range is measurement; Kalman gain balances prediction vs. measurement trust.
Prompting wearers to move is a deliberate design choice: motion makes range-only localization observable where static ranging alone cannot determine bearing.
