Field Trials with Wireless Sensor Networks: Issues and Remedies Volker Turau Matthias Witt Marcus Venzke Hamburg University of Technology, Institute of Telematics Schwarzenbergstraße 95, 21073 Hamburg, Germany Abstract Extensive field trials are essential to evaluate protocols and algorithms for wireless sensor networks. The high costs of such trials demand for a systematic approach. This paper discusses the issues of setting up a field trial such as packaging, logging, and deployment. Details about a concrete field trial to evaluate a topology discovery algorithm are presented. 1. Introduction Wireless sensor networks (WSN) are networks of small, resource-constrained wireless devices equipped with sensors embedded in a dynamic physical environment. The field gained momentum around the year 2000 with the availabil- ity of relatively inexpensive nodes, sensors, and radios [1]. Projects like Smart Dust provided the motivating vision for this research. In principle WSNs are distributed systems, but they pose some unique challenges, foremost resource limitations, high failure rates, and ad hoc deployment. Within the last years many algorithms and protocols have been proposed to solve the problems of WSN. The vast majority of these has not been implemented on real sensor networks, but evaluated using simulation tools. Simulations are a valuable and cheap means to compare specific aspects of different algorithms solving the same problem (e. g., routing or data aggregation), but currently no simulation tool is capable to allow for all imponderabilities of a real deployment of a WSN in a harsh environment over a longer period of time. To attain a deeper insight into WSNs, experiments with real deployments are indispensable. But up to today the number of long-term deployed WSNs is extremely low compared with the number of theoretical results. Extensive field trials require a lot of resources and are very time-consuming. Moreover, systematic approaches for WSN deployment are missing today. While algorithms, protocols, and architectures have been investigated, not much has been achieved in the area of deployment support, pre- deployment debugging, or even a methodology that allows for reliable monitoring of WSNs. This paper discusses the issues of setting up and running a field trial for WSN in general. In the second part details about a concrete field trial to evaluate a topology discovery algorithm for WSNs are presented. A detailed analysis about the results of this trial can be found in [2]. 2. Related Work A driving application area of WSN research is environ- mental monitoring. The vision is to revolutionize biological and climate monitoring, by providing instant data at gran- ularities unrealizable by other means. The majority of field trials of WSNs have this application background. One of the first larger field trials was the habitat-monitoring system on Great Duck Island [3], [4]. The WSN collected data about underground nesting burrows and surface micro-climates such as temperature, humidity, and occupancy. The data was used to correlate nesting patterns with micro-climates. Two different networks were deployed, a 50-node single-hop network and a 100-node multi-hop network covering an area of 200×100 m. Nodes periodically sampled data and routed it to a device equipped with a high-gain antenna to transfer the data to a distant solar-powered laptop with satellite Internet link. Apart from a few parameters that could be tuned in situ, the nodes were completely preconfigured. To verify the collected data, an independent verification network was installed utilizing con- ventional technologies. This network collected short movies using in-burrow cameras for the final verification process. The sensor networks deployed in field trials are often restricted in some way: single-hop network, artificial envi- ronments, very small networks. R¨ omer et al. give a review of several deployments [5]. Apart from primarily application- oriented trials there were also field trials focusing on link stability and network protocols. Woo et al. conducted a trial with a 50-node network placed inside a building [6]. The short experiments were focused on link estimation, neighborhood table management, and reliable routing protocol techniques. EmStar [7] is a software framework for WSN deployment support. It consists of several tools for simulation, emulation, and visualization of applications. Also hybrid modes for combination of simulation and real wireless communication are available. A set of libraries and services exists for plugging them together in custom applications. EmStar is aimed at the high-end spectrum of WSN platforms, namely 32-bit embedded microservers. 3. Setting up a Field Trial Outdoor field trials of WSNs pose problems that do not occur in lab experiments. Some of the reasons are harsh environmental conditions (e. g. low temperatures), long-term unattended operation, complex deployment process, connec- tion to the wired world, and large number of nodes. This section discusses the set up of a field trial of a WSN and evaluates different alternatives.