A New Method for Kalman Filter Tuning

Kalman filter is a well-known estimator. For a target tracking scenario, the filter needs two input parameters. These parameters are called process noise covariance and measurement noise covariance. It is necessary to determine these parameters in order to design a consistent and high-performance target tracker. These parameters vary widely depending on the problem and measurement system. Therefore, there is no commonly accepted method to tune the filter. In this study, we examined the performance and consistency of Kalman filter through a simple target tracking scenario. As a result of the study, three metrics are defined. These metrics can be calculated using measurement data and estimation results only. In the study, it has been shown that Kalman filter can be tuned using this metric. It is not necessary to know the actual states to calculate the metric. This fact makes the method useful for practical applications.