SeeDTV: Deployment-Time Validation for Wireless Sensor Networks H. Liu, L. Selavo, J. Stankovic Department of Computer Science University of Virginia {hl 4d |selavo|stankovic}@cs.virginia.edu Abstract Deployment of a wireless sensor network (WSN) system is a critical step because theoretical models and assumptions often differ from real environmental characteristics and per- formance at the deployment site. In addition, such systems are often located in areas that are difficult to reach or even in- accessible for certain periods of time. Therefore, it is impera- tive to verify the functionality of the system at the time of the deployment, thus lowering the risk of early failures. Coinci- dentally, the validation minimizes the expense of revisiting the site in the near future for re-deployment, maintenance, or repairs. In this paper we present a deployment time valida- tion framework SeeDTV that consists of techniques and pro- cedures for WSN status assesment and verification. SeeDTV is supported by a portable, lightweight, and low power in-situ user interface device SeeMote. SeeDTV has demonstrated the potential for early problem detection at three levels of WSN in-situ validation: sensor node devices, wireless net- work physical and logical integrity, and connectivity to the back-end such as a data server over the Internet. SeeDTV is presented in the context of LUSTER – an environmental sen- sor network for ecological monitoring under a shrub thicket canopy on islands off the coast of Virginia. 1 INTRODUCTION The theory and design of wireless sensor networks (WSN) have progressed admirably leading to deployment of inter- esting and valuable monitoring systems. However, the per- formance of these systems after the deployment often differ considerably from the expected or observed at the time of the development. For example, Szewczyk et al. in their paper present performance of their WSN after the deployment on an island off the coast of Maine [1]. The authors used Woo et al. routing algorithms to measure the packet reception ratio (PRR) performance, and found that the multi-hop networks Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. EmNets’07, June 25–26, 2007, Cork, Ireland Copyright 2007 ACM 978-1-59593-694-3/07/06 ...$5.00 deteriorate over time, with some networks delivering under 30 percent of packets [2]. Similarly, Tolle et al. deployed a WSN on a redwood tree, but a large part of the network failed to form the routing topology [3]. Approximately 15 percent of the deployed nodes completely ceased function- ing in one week because their batteries had been exhausted due to a design flaw [2]. Other evaluations of real deploy- ments reported that reduced transmission ranges, dirty sen- sors and short life-time of nodes did not match the expecta- tions obtained through simulations [4, 5]. Li et al. proposed a software environment, called POWER, for planning and de- ploying wireless sensor networks to real environments to re- duce deployment risks [6]. Deployment Support Network (DSN) provides visibility and control with an additional backbone network [7]. Other application-cooperative man- agement mechanisms include the Sensor Network Manage- ment System (SNMS) [8] and the Sympathy debugging tool for pre-deployment sensor networks [9]. However, POWER does not anticipate the need for low-power in-field user in- terfaces, DSN needs additional nodes leading to higher cost and energy consumption, SNMS is too complicated to be ex- ecuted in inaccessible deployment, and Sympathy can only provide robustness guarantee before the system is deployed. Part of the discrepancy between the expected and real per- formance is because the communication range of a single node is hard to predict. Contrary to the assumption that the range profile is circular, as used in many calculations, it de- pends on the hardware design, the deployment location and the environment, and the communication link quality varies considerably as the batteries become depleted [10]. All this evidence suggests a critical issue in WSN: The deployment of a WSN is a non-trivial task and needs certain techniques and tools to validate the performance and prevent early failures of the system in the new environment. In a certain variety of WSN applications, such as large scale and hard to access systems, none of the existing so- lutions for performance prediction are satisfactory. In this paper we present a deployment-time validation (DTV) ap- proach for wireless sensor networks, called SeeDTV, that consists of techniques and tools to ensure successful deploy- ment of the WSN and assists in maintenance during the sub- sequent visits of the deployment site. The validation targets three levels of the system: per-node validation, communica- tions validation, and the whole system including application specific validation.