Real Time Geolocation Outlier Detection and Noise Correction for Uncertain Route

In this paper, we propose a method to identify outliers in geolocation information in real time, while a consignment is being tracked. We use Chebyshev's inequality to give a weak upper bound on the distance deviation. This upper bound is used to label the first deviation point as an outlier with a 25% chance. This allows us to conclude that the probability of two consecutive points on an alternate route being the outlier is 6.25%. With probability decreasing in successive points, two consecutive outlier points are considered as a deviation from the route. Since identifying deviation is inbuilt into the outlier detection method, we are also able to identify alternate routes in real time. A correction of the non-outlier points is also done using PostGIS and a method called projection to the nearest lane. Finally, we use Gaussian white noise to generate different levels of noisy geolocations. We show that, as the noise increases, it gets harder to identify outliers which allows us to reach a second theoretical upper bound on the level of permissible noise.