Measurement Augmentation to Compensate for Sensor Registration Using a Neural Kalman Filter

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, but standard systems can easily compensate for only some types of errors. There are well known techniques to compensate for errors that result from such issues as noise and sensor accuracy limitations, but other types of errors are not easily compensated for in standard systems. In target tracking, sensor registration, the result of sensor location and orientation self-reference errors, is a type of error that is not easily compensated for because it causes a deterministic bias or parameter drift, rather than random noise error. Previously, a modification of an adaptive tracking technique based upon the neural extended Kalman filter was proposed as a technique to provide for on-line calibration for the sensor models. In this work, that technique has been improved by modifying the input vector of the neural network to make use of a combination of the target and ownship state variables. The result is an ability to correct for errors such as sensors registration using less computational complexity in the neural network than previously required.