While GPS devices are exceptionally great when used to determine your positions on the surface of the planet with a margin of error of about 13 meters, they are not accurate enough to implement them in cars for example, to determine each other's position on the road, in order to improve security and avoid traffic jams. Furthermore, the GPS receivers commonly used can give multiple location errors, because of the geometry of visible satellites, atmospheric interference and clock synchronization error.

Thus, a group of researchers from the University of Waterloo and California Polytechnic State University, working on a project intended to design a lane detection system on the highways, decided that the traditional GPS devices were not accurate enough for such applications, and had to design their own. The new GPS detection system that they've developed uses cheap receivers, linked to a filter that calculates the positions of other cars on the lanes.

Thanh-Son Dao explains that, by using low-cost GPS receivers, the same performance can be obtained in such applications as in more accurate devices. Nevertheless, the absolute position of a vehicle can never be detected accurately, because of measuring errors; however, while trying to determine the relative position between the vehicles, the measuring error becomes ever smaller, enabling an accurate detection.

Receivers used in the experiments work on the basis of the Circular Error Portable or CEP, having measurement accuracies up to 5 meters, meaning the half of the data points are received inside the 5 meter radius around the device, while the other half - outside the circle radius. To remove the parasite signals emitted by multiple GPS receivers, the developers created a filter that screens the signal through Markov probabilistic distribution localization algorithm, thus enhancing the systems sensibility.

This filtering system also enables a simple detection method, by determining the probability that the vehicle has changed lanes, instead of determining the actual position of a particular vehicle on a lane. The algorithm predicts a path by using at least two previous position estimates, which it compares with the current positions of the vehicles. If it detects a change in the predicted position and the current position, most likely the vehicle is passing through a curve on the highway, or it has changed lanes. By measuring the relative distances between the predicted position and the current one, the system decides which one of the two possible outcomes has taken place. If the distance is greater than that measured during changing lanes, then probably the car is passing through an arc of the road.

During real world testings, they have successfully demonstrated that the system has an accuracy well over 90 percent, however, the problem related to the eventual loss of satellite signal still remains. Dao argues that situations like passing through a long tunnel where satellite signal in unavailable could be eliminated with the help of inertial instruments, linked to the GPS units.