Estimating GPS/ADS-B Dropouts in the OpenSky Network Platform

Unmanned Aerial Systems (UAS) are used for many purposes. Their navigation systems can be compromised in a variety of ways, creating potentially hazardous situations. The integrity of navigation systems, such as Automatic Dependent Surveillance Broadcast (ADS-B) and Global Positioning System (GPS), must be analyzed to detect anything that threatens their integrity. Through analysis of OpenSky Network data, the separation of individual aircraft and individual flight paths was achieved. An analysis of each flight using statistical methods allowed data to be scored and categorized as a dropout, noise, normal, or erroneous packet. Machine learning models were applied to predict these labels using only latitude, longitude, velocity, geo-altitude, and dropout length. The Random Forest Classifier model performed extremely well ( 98%), followed by the K-Nearest Neighbor clustering model ∼which had poor accuracy (∼38%) but provided insight into data locality, and finally a Linear Support Vector Classifier that performed poorly (∼51%).