Experimental Study for Multi-layer Parameter Configuration of WSN Links

Many applications of wireless sensor networks (WSNs) need to balance multiple yet often conflicting performance requirements such as high energy efficiency, high throughput, low delay and low loss. Finding appropriate WSN parameter configuration to achieve the best trade-off requires in depth understanding of the joint effect of key parameters residing at different layers on the performance. In this paper, we present an extensive experimental study on the data delivery performance of aWSN link, where 4 major performance metrics, namely energy, throughput, delay and loss, were measured over 6 months under around 50 thousand parameter configurations of 7 key stack parameters. Different from existing work, rich observations are made out of the extensive measurement data, with the focus on the joint effect of these parameters on the performance. Specifically, for each of the four performance metrics, a set of guidelines is derived for parameter optimization. In addition, we propose empirical models for each performance metric to quantify the joint effects, which enable finding optimal settings for parameters such as payload size or retransmissions, in consideration of link quality and other parameter settings, to achieve better performance trade-offs. To demonstrate the potential of this work, the obtained joint parameter optimization results are applied to an example. The outcome is compared with those achieved by following representative single-parameter tuning guidelines from the literature. The comparison reveals that by considering the joint effect of multi-layer parameters together, a WSN application can obtain a much improved performance trade-off.