Dynamic Uncertainty
Modeling errors in a state estimation system that scale with the magnitude of the system’s motion demonstrate the difficulty of tracking high-speed maneuvers. Incorporating rate-dependent process noise allows a filter to increase its internal uncertainty during rapid turns or accelerations. This prevents the filter from becoming overconfident in its motion model.
Adaptive Tuning
The noise covariance matrix is adjusted dynamically based on the current state estimates. When using rate-dependent process noise, the system assumes that higher velocities lead to larger model discrepancies. This is common in aerodynamic models where drag and turbulence are non-linear.
Tracking Benefit
During steady-state flight, the process noise is kept low to allow for smooth estimation. However, when the vehicle initiates a sharp turn, rate-dependent process noise expands the search window for the next state. This allows the filter to follow the maneuver without lagging behind the actual position.
It also prevents the innovation covariance from becoming too small, which would lead to the rejection of valid sensor data. The transition between these states must be smooth to avoid numerical instability. Calibration of the scaling factor requires extensive testing against ground-truth data.
Filter Resilience
This approach is particularly useful for tracking targets with unpredictable motion patterns. Implementing rate-dependent process noise ensures the estimator remains stable during critical mission phases.
