
An algorithm that estimates the state of a system from noisy measurements over time.The Kalman filter is an algorithm that produces estimates of a system's true state from a series of measurements that are corrupted by noise. It does this by combining a prediction of where the system should be (based on a physical model) with the actual measurements, weighting each according to their uncertainty. The result is an optimal estimate that minimizes the mean squared error, provided the noise is Gaussian and the system dynamics are linear.
Developed by Rudolf E. Kálmán in 1960, building on earlier work by others on Wiener filtering and state-space models.
The filter operates in two steps: predict and update. In the predict step, it uses the system's dynamics to project the current state forward in time, along with an estimate of the uncertainty (covariance) of that projection. In the update step, it compares the new measurement to the prediction, and computes a weighted average — the Kalman gain — that balances trust in the model versus trust in the sensor. The gain is chosen to minimize the posterior uncertainty, and the process repeats at each time step.
Used in navigation systems (GPS, inertial guidance), robotics, computer vision, and signal processing; in the cited articles it appears in contexts like lidar data analysis and quantum state reconstruction.
The Kalman filter was used in the Apollo program's navigation computer to estimate the spacecraft's position during lunar missions.





