Understanding the Kalman Filter with a Simple Radar Example
5 hours ago
- #predict-update
- #filter-algorithm
- #state-estimation
- The Kalman Filter is an algorithm for state estimation and prediction under uncertainty, used in fields like tracking, navigation, robotics, and finance.
- It operates in a predict-update loop: prediction extrapolates the state using a dynamic model, and update combines prediction with new measurements weighted by uncertainty.
- Key equations include state extrapolation (using state transition matrix F), covariance extrapolation (with process noise Q), state update (with Kalman gain K), and covariance update.
- The Kalman gain optimally weights prediction and measurement based on their uncertainties to minimize estimation error.
- Example demonstrated a radar tracking aircraft: initialization used first measurement, prediction estimated next state, and update fused noisy second measurement to reduce uncertainty.
- Process noise accounts for model inaccuracies (e.g., wind), while measurement noise reflects sensor errors; covariance matrices quantify these uncertainties.
- The filter reduces estimation uncertainty by combining information, even from noisy measurements, and is optimal under assumed linear Gaussian conditions.