An engineer does not often describe an electro-mechanical system as "more than the sum of its parts." But Assistant Professor Demoz Gebre-Egziabher of the Department of Aerospace Engineering and Mechanics sees the combination of two relatively well known guidance technologies forming a system which delivers better performance than either technology would be capable of by itself.

Knowing precisely where a vehicle is located is one of the core requirements of many intelligent transportation systems applications. To locate a target in space relative to a coordinate system, two main approaches are possible:

- The first is the "dead reckoning" method of computing a moving target's position relative to its starting point by keeping track of the direction and speed of its movement; inertial navigation systems, which use accelerometers, gyroscopes, and magnetic compasses to measure changes in speed and direction, are a common example. The accuracy of these systems is determined by the sensitivity of their internal components.
- The second method is "positional" navigation: continually refiguring the spatial relationships between the target and several fixed landmarks; this method is exemplified by satellite guidance systems like GPS or the European Galileo network. The accuracy of these systems is determined by the precision of their measurements of the distance to external landmarks.

At the ITS Institute, high-accuracy GPS applications are among the most visible research topics. (See previous seminars on GPS/navigation). Positional navigation system like GPS are capable of excellent accuracy (on the order of two centimeters using a differential correction signal), but they suffer from certain limitations inherent in the system design. For example, a clear line of sight between the receiver and multiple satellites or beacons is required. Disrupting the signal path by passing under a bridge or between tall buildings means literally "flying blind" until the signal can be reacquired.

In contrast, inertial navigation systems don't need any outside reference points, and function equally well underground as in the open air. Also, unlike positional systems, they provide location information at a very high rate because they don't need to reference external sources. Their drawback is that errors in their measurements build up rapidly over time, because each new position computation depends on the results of the one before. The probability that an inertial measurement of position is accurate therefore decreases rapidly over time unless the system can be recalibrated against an external positional reference.

Given the relative strengths and weaknesses of each type of navigation system, is it possible to combine the two in such a way that each will cancel out the other's shortcomings? Gebre-Egziabher's current work focuses on techniques for accomplishing this goal, especially the development of algorithms that combine data from an inertial and a positional system to generate correction data for the system as a whole.

Such an algorithm would essentially "factor out" errors based on comparison of data from different sources and a knowledge of vehicle dynamics. For a problem in two dimensions, such as driving a car, the various components of an integrated system produce no fewer than eight factors subject to error. For a three-dimensional problem, such as landing an airplane, the number of error-prone factors increases.

Fortunately, the field of control systems theory has produced mathematical tools for removing errors from multidimensional systems. By constructing an "observability matrix," it is possible to determine which errors can be eliminated by calculation. But the fact that the vehicle's movements affect which errors can be eliminated at a given instant in time makes the problem more than a simple exercise in mechanics.

Gebre-Egziabher and his students are currently at work on multiple projects relating to the integration of inertial and positional navigation systems. These include an integrated inertial/GPS navigation system for micro-UAVs ("µ-UAVs")—very small Unmanned Aerial Vehicles that have many potential civilian and military applications. Their prototype currently fits in the trunk of a Toyota, but the ultimate goal is to create a vehicle "smaller than an Altoids can," says Gebre-Egziabher.

Another current project involves improving the robustness of GPS navigation systems by adding the ability to calculate the Doppler shift of satellite signals using an onboard inertial navigation component. This would enhance the ability of GPS-based navigation systems to overcome the effects of radio-frequency interference.