Online Sensor Modeling Using a Neural Kalman Filter

Abstract

Sensor-measurement systems rely upon knowledge of the functional dynamics between system states and the measured outputs. Errors in sensor measurements come from a variety of sources. While there are well-known techniques to compensate for those errors that result from such issues as noise and sensor-accuracy limitations, other types, such as those that are more deterministic, can result in biases that are not easily compensated for in standard systems. A modification of an adaptive tracking technique based on the neural extended Kalman filter is proposed as a technique to provide for online calibration for the sensor models. Previously, the technique has been applied to tracking problems and successfully improved the motion model of a target when a maneuver occurs. In this new application of the technique, the sensor dynamics are learned rather than the target dynamics.