An Efficient Polling Strategy for Bluetooth Piconets using Channel State Information João Henrique Kleinschmidt, Marcelo E. Pellenz and Luiz A. de Paula Lima Jr. Graduate Program in Computer Science Pontifical Catholic University of Paraná Curitiba – PR, Brazil {joaohk, marcelo, laplima}@ppgia.pucpr.br Abstract - Bluetooth is an emerging technology for personal wireless communications and is being applied in many scenarios. In this paper we investigate the data scheduling policy for Bluetooth piconets. In general the performance evaluation of most piconet scheduling algorithms recently proposed do not consider the packet losses in the wireless channel. However in environments with high interference or mobility, the channel conditions can mitigate the performance of the polling strategies. In this work we propose a new piconet scheduling algorithm based on the estimated channel state information. The channel is modeled using the Nakagami-m distribution. The channel quality estimation is based on the received signal-to-noise ratio and on the Nakagami fading parameter m. These two metrics are used by the scheduling algorithm to define the best polling strategy. Simulation results demonstrate the good algorithm performance for different traffic conditions. Keywords — Bluetooth, Nakagami-m fading, scheduling, wireless networks. I. INTRODUCTION Bluetooth is emerging as an important standard [1] for short range and low-power wireless communications. It operates in the 2.4 GHz ISM (Industrial, Scientific and Medical) band employing a frequency-hopping spread spectrum technique. The transmission rate is up to 1 Mbps, using GFSK (Gaussian Frequency Shift Keying) modulation. The Bluetooth MAC protocol is designed to facilitate the construction of ad hoc networks. The devices can communicate with each other forming a network with up to eight nodes, called piconet. Within a piconet, one device is assigned as a master node and the others devices act as slave nodes. Devices in different piconets can communicate using a structure called scatternet. The channel is divided in time slots of 625 μs. A time-division duplex (TDD) scheme is used for full-duplex operation. For data transmission Bluetooth employs seven asynchronous packet types. Each packet may occupy 1, 3 or 5 time slots. The throughput of Bluetooth links using asynchronous packets was investigated in [2] for the additive white Gaussian noise (AWGN) channel and for the Rayleigh fading channel. In [3], we extended the results presented in [2] looking into the performance of Bluetooth links in Nakagami-m fading channels. The Nakagami-m distribution [4] allows a better characterization of real channels because it spans, via the parameter m, the widest range of multipath fading distributions. For m=1 we get the Rayleigh distribution. Using m<1 or m>1 we obtain fading intensities more and less severe than Rayleigh, respectively. The packet error rates (PER) of the asynchronous packets for different values of m and signal-to-noise ratios are derived in [3]. In this work we apply the channel modeling derived in [3] for the Nakagami fading channel in order to study the new proposed scheduling algorithm, which uses the channel quality information in the scheduling policy. This paper is structured as follows: in Section 2 some issues about piconet scheduling and related works are presented. In Section 3 we draw some considerations about channel state estimation. Section 4 proposes a new strategy based on channel quality estimation and Section 5 shows the simulation results for different scenarios. Finally, conclusions are drawn in Section 6. II. RELATED WORK ON PICONET SCHEDULING In a Bluetooth piconet, the channel access is controlled by the master. A slave can send a packet only if it receives a polling packet from the master. The master transmits packets to the slave in even slots while the slave transmits packets to the master in odd slots. Thus, Bluetooth is a master driven TDD standard and this poses several challenges in scheduling algorithms since there could be a waste of slots if only the master or the slave has data to send. Recently, many schemes have been proposed in the literature for piconet and scatternet scheduling. In [5], several polling schemes are compared. In the round robin scheme a fixed cyclic order is defined and a single chance to transmit is given to each master-slave queue pair. The exhaustive round robin (ERR) also uses a fixed order but the master does not switch to the next slave until both the master and the slave queues are empty. The main disadvantage of the ERR is that the channel can be captured by stations generating traffic higher than the system capacity. A limited round robin (LRR) scheme that limits the number t of transmissions can solves this problem. A new scheme called LWRR (limited and weighted round robin) with weights dynamically changed according to the observed queue status is also presented in [5]. Other works about piconet scheduling consider QoS issues in Bluetooth, such as [6] and [7]. The results in [5], [6] and [7] do not consider any loss model for the wireless channels.