Advanced Filtering: Unscented Kalman vs Standard Kalman

When Standard Kalman Falls Short...

The Standard Kalman Filter (KF) works on the assumption that the system moves linearly (e.g., a car moving at constant speed). However, the real world is chaotic.

Extended (EKF) and Unscented Kalman (UKF)

If the system is non-linear (e.g., complex joint movements of a robot arm or a drone hit by sudden wind), standard KF produces erroneous results.

1. Extended Kalman Filter (EKF): Tries to "force" the system into linearity using mathematical derivatives (Jacobian matrices). It requires processing power and can sometimes become unstable.
2. Unscented Kalman Filter (UKF): Instead of linearizing the system, it takes a sample of possible scenarios (sigma points) and simulates them. Since there is no derivative calculation, it is more stable and usually yields more accurate results than EKF.

Amazeng's AI-powered modules utilize lightweight UKF algorithms to clean complex signal noise.