Bluetooth scatternet formation from a time-efficiency perspective

The Bluetooth Scatternet Formation (BSF) problem consists of interconnecting piconets in order to form a multi-hop topology. While a large number of BSF algorithms have been proposed, only few address time as a key parameter, and when doing so, virtually none of the solutions were tested under realistic settings. In particular, the baseband and link layers of Bluetooth are highly specific and known to have crucial impacts on performance. In this paper, we revisit performance studies for a number of time-efficient BSF algorithms, focusing on BlueStars, BlueMesh, and BlueMIS. We also introduce a novel time-efficient BSF algorithm called BSF-UED (for BSF based on Unnecessary-Edges Deletion), which forms connected scatternets deterministically and limits the outdegree of nodes to 7 heuristically. The performance of the algorithm is evaluated through detailed simulation experiments that take into account the low-level specificities of Bluetooth. We show that BSF-UED compares favorably against BlueMesh while requiring only 1/3 of its execution time. Only BlueStars is faster than BSF-UED, but at the cost of a very large number of slaves per master (much more than 7), which makes it impractical in many scenarios.