Affiliations: [a] SITE, University of Ottawa, Ottawa, Ontario K1N 6N5, Canada. E-mail: zaguia@site.uottawa.ca, ydaadaa@site.uottawa.ca | [b] Electronic, Electrical & Computer Engineering, The University of Birmingham, Birmingham, UK. E-mail: stojmenovic@storm.ca
Abstract: Bluetooth standard allows the creation of piconets, with one node serving as its master and up to seven nodes serving as slaves. A Bluetooth ad hoc network can also be formed by interconnecting several piconets into a scatternet. Given a set of Bluetooth nodes which are positioned so that their unit disk graph is connected, the Bluetooth scatternet formation (BSF) problem is to select piconets, and assign master and slave roles in each piconet, so that the obtained scatternet is connected, has some desirable properties and good performance with respect to some metrics. In this article we propose BlueMis, a new BSF protocol based on maximal independent sets. BlueMis is a two phase protocol, in which the first phase is discovery and the second phase, is scatternet formation. In the discovery process, two hop neighbors are discovered. The second phase of BlueMis uses two iterations. In the first iteration a piconet containing a maximal independent set is constructed for every device, while the second iteration attempts to simplify the scatternet structure and to delete piconets not essential for the connectivity. This novel protocol is an attempt to simplify the BlueMesh procedure. Simulation results show that the average number of iterations in BlueMesh ranges from 2.7 to 4.5, compared to only 2 for BlueMis. The reduced overhead is a major gain and it represents our primary contribution in this paper. BlueMis is implemented and compared with BlueMesh, in terms of various characteristics. Experiments show that BlueMesh performs slightly better, in terms of the number of piconets and the number of slaves per piconet.
