Desensitized Optimal Filtering and Sensor Fusion Tool Kit, Phase I

It is proposed to develop desensitized optimal filtering techniques and to implement these algorithms in a navigation and sensor fusion tool kit. These proposed desensitized optimal filtering techniques include recent advances in robust and/or adaptive generalized Kalman and Sigma-Point filters for non-Gaussian problems with uncertain error statistics, as well as a proposed new technique to desensitize the Kalman filter with respect to parameter uncertainties using a robust trajectory optimization approach called Desensitized Optimal Control. These techniques will be implemented in a relatively generic environment which enables the user to import dynamics and measurement models necessary to apply these filtering techniques to a particular navigation and sensor fusion problem. A variety of sensor models and noise distributions will be available for the user to select, and Monte-Carlo analysis capability will be built into the tool kit to enable statistical performance evaluations. The tool kit will also have a modularized structure so that the modules can be readily integrated with other applications. More »