An efficient reconstruction approach for improving Bluetree scatternet formation in personal area networks

Abstract

Bluetooth is a communication technology for personal area networks (PANs). To support communication with more than eight bluetooth-enabled devices, a scatternet must be formed in the PAN. Bluetree is one commonly used topology for scatternet formation. To reduce traffic load of the Bluetree scatternet, we use the piconet transfer concept to move piconets on a well-formed Bluetree scatternet. The piconet movements are performed based on a distributed manner using two well-known tree traversal procedures: post-order traversal and level-order traversal. These two procedures do not take much computation time, where time complexities are O(n_p) and n_p is the number of piconets on a Bluetree scatternet (not the number of nodes on a Bluetree). Compared with previous approaches, the proposed approach can greatly reduce the traffic load and computational costs. Finally, simulation experiments show the effectiveness of the proposed approach in improving the formation of Bluetree scatternet.

Introduction

Bluetooth is a low-cost, short range wireless communication technology for personal area networks (PANs), standardized as IEEE 802.15.1 (Bluetooth, 2006). Bluetooth-enabled devices such as cellphones or PDAs can be connected to form several small networks called piconets. In a piconet, there are at most eight devices, one of which takes the role of the master, the rest, slaves. To connect more than eight devices, piconets can be further interconnected by designating some devices as bridges to form a larger network called a scatternet. In current Bluetooth standards, piconet formation is a well-defined process. However, scatternet formation lacks clear specifications. As a result, many researchers have explored scatternet formation.

In the existing literature on scatternet formation, the scatternet is formed into one of the following topologies: BlueMesh, Bluering, BlueStar, and Bluetree (Persson et al., 2005; McDermott-Wells, 2004–2005). Bluetree is a commonly used topology that enables link scheduling and packet routing to be easily performed in a scatternet. The formation of Bluetree scatternet is first mentioned in Zaruba et al. (2001). However, as noted in Tan et al. (2002), the formation algorithm of Zaruba et al. (2001) does not describe how to perform node discovery and link establishment to establish and maintain a Bluetree scatternet. In Tan et al. (2002) and Cuomo et al. (2003), two detailed algorithms for forming a Bluetree scatternet are given: Tree Scatternet Formation (TSF; Tan et al., 2002) and Self-Healing Algorithm Producing multi-hop Bluetooth scattERnets (SHAPER; Cuomo et al., 2003). SHAPER extends TSF to release the assumption of the single-hop scenario. In SHAPER, if all nodes cannot directly communicate with each other, this algorithm can still form a Bluetree scatternet in a multi-hop scenario. However, in the above two algorithms, traffic demand between piconets is not considered in the formation of the Bluetree scatternet.

Thus, the main goal of this research is to re-construct a well-formed Bluetree scatternet that takes the traffic load into consideration. Similar issues have been discussed in Tekkalmaz et al. (2006) and Chang and Chang (2007). In the approach of Tekkalmaz et al. (2006), several techniques are used to achieve reduction in traffic load: node transfer, piconet merge, role reassignment, and piconet division. Node transfer and piconet merge are used to put nodes with high traffic between them in the same piconet. The role assignment then re-selects appropriate nodes to act as the master or bridge. Due to the need to calculate all possible selection cases, the approach of Tekkalmaz et al. (2006) results in complex computations in role assignment. While piconet division can improve total bandwidth usage in a scatternet by splitting a piconet into two piconets, the average route length may be increased. The approach of Chang and Chang (2007) also uses node transfer and role switch (role reassignment) to change the topology of a formed scatternet to reduce the traffic load. However, node transfer is executed when two neighboring piconets have a large interpiconet traffic load. Role switch in the approach of Chang and Chang (2007) also introduces a time-consuming matrix computation. As noted in Roy et al. (2007), since a Bluetooth node has constraints in both power and memory, running computation-intensive or space-consuming algorithms is not feasible.

In our approach, tree traversal and piconet transfer techniques are used to reduce the traffic load of a Bluetree scatternet. Tree traversal is used to systematically assist each piconet in taking a small computation time to determine its best location on the Bluetree scatternet and then move there. Piconet transfer can aggregate a greater number of traffic flows for handling in combination than node transfer. When there are a large number of traffic flows in the Bluetree scatternet, the proposed approach has better scalability. Finally, we perform simulations to show the effectiveness of the proposed approach in terms of the following performance metrics: traffic cost reduction, transmission energy reduction, computation cost improvement, message overhead improvement, and scalability improvement for handling the traffic flows.

The rest of the paper is organized as follows. Section 2 gives preliminaries of this paper. Section 3 proposes our approach. Section 4 evaluates the performance of the proposed approach. Finally, conclusions are presented in Section 5.