Detecting Outlier Positions in Publicly Available ADS-B Data Sets

Automatic Dependent Surveillance - Broadcast (ADS-B) is one of the most widely used surveillance technologies in aviation. The data from ADS-Bare extensively used for commercial and research purposes. Incorrect positional messages present in the data could lead to wrong or misleading scientific conclusions. The paper presents two different methods for ADS-B position outlier detection. First, the use of Density Based Spatial Clustering of Applications with Noise algorithm is proposed. Second, the use of the Interacting Multiple Model algorithm is proposed, where the statistical distribution of the innovation is as a means for outlier detection. Both algorithms are compared on a large publicly available data set consisting of 513 million messages. The advantages and disadvantages of both of the algorithms are discussed and recommendations for their use are provided. The results indicate that up to 0.75 % of ADS-B Airborne Position Messages should be discarded.