An optimized filter architecture incorporating a neural net

The Kalman filtering process is generally a combination of predicting the state variables and then correcting them with the measurement data from the sensor systems. The approach presented here is to use a neural network to make the corrections to the state variable estimates within a Kalman filter structure. A model is used for the prediction of the state variable estimates and corrections are made to the state estimates based on transformation of the measurement residuals. A key assumption in this filter concept is that the correction to the state estimate is a function of the measurement residuals for all practical applications of filters. This approach uses the function approximation capabilities of some neural network architectures in combination with well established filter theory. The concept provides the potential for improved accuracy and increased robustness in filter applications