Temperature strongly affects the operation of integrated circuits, and its impact has been largely investigated on a device level. However, the impact of temperature variations on networks of multiple devices is far less understood and requires investigation. We aim to close this gap and analyse the impact of temperature fluctuations on low-power wireless sensor networks, a key enabling technology of pervasive computing. As we are moving forward into an era of human-centric safety-critical applications (e.g., smart health and intelligent transportation systems), it is particularly important to make sure that a networked system offers a reliable and deterministic performance despite all possible temperature changes over its deployment lifetime. In this demo, we present a testbed infrastructure based on infra-red heating lamps that allows to vary the on-board temperature of sensor nodes on a large scale in a repeatable fashion. Using this experimental infrastructure, we show the effects of temperature variations on network performance in two different ways. First, in a small-scale local testbed at PerCom, we highlight the degradation of the wireless link quality at high temperatures, and show that the performance of radio transceivers is temperature-dependent. We quantify this degradation and parametrize the dependency between temperature and link quality using the signal strength information captured between four wireless sensor nodes. Second, we connect remotely to our large-scale experimental infrastructure at TU Graz, and assess the impact of temperature variations on the performance of state-of-the-art network protocols, showing that the typical outdoor temperature fluctuations occurring during 24-hours do affect key network metrics such as throughput, delay, and lifetime.

Temperature has a strong impact on the operations of all electrical and electronic components. In wireless sensor nodes, temperature variations can lead to loss of synchronization, degradation of the link quality, or early battery depletion, and can therefore affect key network metrics such as throughput, delay, and lifetime. Considering that most outdoor deployments are exposed to strong temperature variations across time and space, a deep understanding of how temperature affects network protocols is fundamental to comprehend flaws in their design and to improve their performance. Existing testbed infrastructures, however, do not allow to systematically study the impact of temperature on wireless sensor networks. In this paper we present TempLab, an extension for wireless sensor network testbeds that allows to control the on-board temperature of sensor nodes and to study the effects of temperature variations on the network performance in a precise and repeatable fashion. TempLab can accurately reproduce traces recorded in outdoor environments with fine granularity, while minimizing the hardware costs and configuration overhead. We use TempLab to analyse the detrimental effects of temperature variations (i) on processing performance, (ii) on a tree routing protocol, and (iii) on CSMA-based MAC protocols, deriving insights that would have not been revealed using existing testbed installations.