Also may use Sherman-Morrison-Woodbury Formula

Definition

The Kalman Filter is an iterative algorithm that estimates the state of a linear dynamic system from noisy measurements. It uses a combination of predictions from a system model and measurements to produce optimal estimates of the state in the least-squares sense.

Equations

1. System Model

A linear system is represented as:

  • State Transition:

    where:

    • : State vector at time .
    • : State transition matrix.
    • : Control matrix.
    • : Control input.
    • : Process noise, .
  • Measurement Equation:

    where:

    • : Measurement vector.
    • : Observation matrix.
    • : Measurement noise, .

Steps in the Kalman Filter

1. Prediction Step

  • State Prediction:
  • Error Covariance Prediction:

2. Update Step

  • Kalman Gain:
  • State Update:
  • Error Covariance Update:

Key Concepts

  1. Prediction:
    Propagates the current state estimate forward in time using the system model.

  2. Correction:
    Adjusts the predicted state using measurements and the Kalman gain to minimize estimation error.

  3. Kalman Gain:
    Balances trust between the prediction (model) and the measurement.

  4. Optimality:
    The Kalman Filter minimizes the mean squared error under the assumption of linearity and Gaussian noise.

Examples

1. One-Dimensional Example (Temperature Sensor)

  • System dynamics: , where .
  • Measurement: , where .

2. 2D Object Tracking

  • State: (position and velocity in 2D).
  • Transition Matrix:
  • Observation: .