We propose to do specific Data Fusion research in the areas of navigation and control. Precise navigation with high update rates is essential for automated vehicle control. Individual sensors currently available cannot meet both requirements. However, the outputs of various sensors can be optimally blended with a Kalman filter for determination of absolute or relative position. Also, given data from external, high accuracy reference systems, internal sensor errors such as biases, scale factor errors, misalignment, etc., can be corrected. Eventually, in the discrete case, the high update rate and high precision required for guidance can be achieved. Generic algorithms for data processing can be developed to manage all the mentioned above issues. The proposed generic nature of the mathematical modeling and software development work will ensure that the inevitable accomodation of new sensors or sensor systems will require minimal changes. The generic aspect will be provided through inheritance of elementary models from the bottom up. This work will have direct application to other areas of data acquisition and control.
The high cost of modifications to infrastructure to provide absolute position information to automated vehicles suggest a hybrid scheme where absolute position beacons provide periodic "fixes" while the vehicles rely on internal, inertial sensors between beacons. It has been suggested that satellite GPS systems might provide positional information alone. All of the currently available GPS receivers have an accuracy vs sample rate trade-off. That ratio is two orders of magnitude away from a number that would be useful in AVCS. In short, GPS alone cannot meet the AVCS navigational needs. We propose to investigate inertial sensors, positional beacon systems, and filtering techniques that may provide a solution to the problem of vehicle position for navigation, maneuver, and control. For example, a $20 mircomachine accelerometer can give several minutes of precise positional information before drift becomes a problem and an external fix is required. This implies that roadside beacons might be several kilometers apart or that a super-accurate GPS fix could be provided every minute.
Automotive application of GPS requires the most accuracy and the most reliability to operate in "fringe" conditions for GPS. GPS reception is easy on an interstate, however under the foliage or in urban canyons, obtaining reliable performance is quite challenging. More channels have to be used to allow the reciver to track all practically visible satellites, holding on to strong signals as the weaker or blocked signals fade out. The task of providing kinematic accuracies less than a centimeter is the subject of advanced research of the leading GPS manufacturing companies. Extensive survey of the available technology reveal that only the two best receivers out of 275 listed in 1995 offer desired performance. Even those two selected are under final development stages by their OEMs. These high accuracy receivers, can be integrated with the micromachined inertial sensor system to provide reliable navigation for AVCS applications. The price for the above mentioned receivers is in the range of $30,000, however, there is a strong indication that it will drop significantly in the next 5 years, as more companies enter the high accuracy competition.