Density-Based Clustering and Performance Enhancement of Aeronautical Ad Hoc Networks

Abstract

In-Flight Entertainment and Connectivity (IFEC) is becoming a key trend and an essential need. A grand challenge is to provide in-flight connectivity in high altitudes, and particularly in isolated locations, such as the oceans, where establishing an air-to-ground link is not possible. Aeronautical Ad-Hoc Networking (AANET) intends to cope with this challenge by forming a network of airplanes having air-to-air (A2A) connections. However, the dynamic nature of such a network is likely to lead to unstable connections. The primary cause of the majority of these stability issues is known to be poor clustering of aircrafts. Consequently, concentrating on aircraft clustering and making them more stable can improve connection. This paper aims to unveil the benefits of density-based clustering to improve the AANET performance. To do so, the paper employs a multi-feature DBSCAN algorithm for the clustering problem that exploits several features of real flight datasets, including latitude, longitude, altitude, direction, and velocity. Instead of a typical distance metric such as Euclidean or Haversine, the technique produces a precomputed distance matrix and feeds it to DBSCAN. This method also includes a weighted scheme to reflect the relative importance of each component of the distance calculation. Simulations under OMNET++ by using real-time flight data point out that packet delivery ratio and end-to-end latency of the state of the art clustering-based AANET solutions can be improved by 40 % and 30 %, respectively. Furthermore, the proposed method achieves a 20% reduction in cluster changes and the number of clusters.