Sasha Slijepcevic

Power efficient organization of wireless sensor networks

Wireless sensor networks have emerged recently as an effective way of monitoring remote or inhospitable physical environments. One of the major challenges in devising such networks lies in the constrained energy and computational resources available to sensor nodes. These constraints must be taken into account at all levels of the system hierarchy. The deployment of sensor nodes is the first step in establishing a sensor network. Since sensor networks contain a large number of sensor nodes, the nodes must be deployed in clusters, where the location of each particular node cannot be fully guaranteed a priori. Therefore, the number of nodes that must be deployed in order to completely cover the whole monitored area is often higher than if a deterministic procedure were used. In networks with stochastically placed nodes, activating only the necessary number of sensor nodes at any particular moment can save energy. We introduce a heuristic that selects mutually exclusive sets of sensor nodes, where the members of each of those sets together completely cover the monitored area. The intervals of activity are the same for all sets, and only one of the sets is active at any time. The experimental results demonstrate that by using only a subset of sensor nodes at each moment, we achieve a significant energy savings while fully preserving coverage.

Localized algorithms in wireless ad-hoc networks: location discovery and sensor exposure

The development of practical, localization algorithms is probably the most needed and most challenging task in wireless ad-hoc sensor networks (WASNs). Localized algorithms are a special type of distributed algorithms where only a subset of nodes in the WASN participate in sensing, communication, and computation. We have developed a generic localized algorithm for solving optimization problems in wireless ad-hoc networks that has five components: (i) data acquisition mechanism, (ii) optimization mechanism, (iii) search expansion rules, (iv) bounding conditions and (v) termination rules. the main idea is to request and process data only locally and only from nodes who are likely to contribute to rapid formation of the final solution. The approach enables two types of optimization: The first, guarantees the fraction of nodes that are contacted while optimizing for solution quality. The second, provides guarantees on solution qualities while minimizing the number of nodes that are contacted and/or amount of communication. The localized optimization approach is applied to two fundamental problems in sensor networks: location discovery and exposure-based coverage. We demonstrate its effective-ness on a number of examples

On communication security in wireless ad-hoc sensor networks

Networks of wireless microsensors for monitoring physical environments have emerged as an important new application area for wireless technology. Key attributes of these new types of networked systems are the severely constrained computational and energy resources, and an ad hoc operational environment. This paper is a study of the communication security aspects of these networks. Resource limitations and specific architecture of sensor networks call for customized security mechanisms. Our approach is to classify the types of data existing in sensor networks, and identify possible communication security threats according to that classification. We propose a communication security scheme where for each type of data we define a corresponding security mechanism. By employing this multitiered security architecture where each mechanism has different resource requirements, we allow for efficient resource management, which is essential for wireless sensor networks.

Location errors in wireless embedded sensor networks: sources, models, and effects on applications

Wireless sensor networks monitor the physical world by taking measurements of physical phenomena. Those measurements, and consequently the results computed from the measurements, may be significantly inaccurate. Therefore, in order to properly design and use wireless sensor networks, one must develop methods that take error sources, error propagation through optimization software, and ultimately their impact on applications, into consideration. In this paper, we focus on location discovery induced errors. We have selected location discovery as the object of our case study since essentially all sensor network computation and communication tasks are dependent on geographical node location data. First, we model the error in input parameters of the location discovery process. Then, we study the impact of errors on three selected applications: exposure, best- and worst-case coverage, and shortest path routing. Furthermore, we examine how the choice of a specific objective function optimized during the location discovery process impacts the errors in results of different applications.

Characterization of location error in wireless sensor networks: analysis and applications

One task in which inaccurate measurements are often used is location discovery, a process where the nodes in a network determine their locations. We have focused on location discovery as the primary target of our study since many sensor network tasks are dependent on location information. We demonstrate the benefits of location error analysis for system software and applications in wireless sensor networks. The technical highlight of our work is a statistically validated parameterized model of location errors that can be used to evaluate the impact of a location discovery algorithm on subsequent tasks. We prove that the distribution of location error can be approximated with a family of Weibull distributions. Then, we show that while performing the location discovery task, the nodes in a network can estimate the parameters of the distribution. Finally, we describe how applications can use the estimated statistical parameters to: (i) estimate the confidence intervals for their results, (ii) organize resource consumption to achieve optimal results in presence of estimated magnitude of error.

Lack of the Golden Standard and Missing Measurements

We propose a new error modeling and optimization-based localization approach for sensor networks in presence of range measurement noise. The approach is solely based on the concept of consistency. The error models are constructed using nonparametric statistical techniques; they do not only indicate the most likely error but also provide the likelihood distribution of particular errors occurring. The models are evaluated using the learn-and-test method and served as the OFs for the task of localization. In addition, we also present a localized localization algorithm where a specified communication cost or the location accuracy is guaranteed while optimizing the other. We evaluate the approach (1) in both GPS-based and GPS-less scenarios; (2) in both centralized and localized manners; (3) in 1D, 2D, and 3D spaces; and (4) in the case when error models are not available a priori, on sets of acoustic ranging-based distance measurements recorded by actual deployed sensor networks. The experimental evaluation indicates that localization of only a few centimeters is consistently achieved when the average and median distance measurement errors are more than a meter, even when the nodes have a low connectivity. Furthermore, we compare the relative performance in terms of location accuracy with several state-of-the-art localization approaches. Finally, several insightful implications about the required conditions for accurate Location Discovery are concluded by analyzing the experimental results.

Future Research Directions

We addressed several well-known and new canonical Location Discovery problems in wireless sensor networks, including engineering change-based node addition, and design and operation of Location Discovery Infrastructure for mobile users. Location Discovery is a generic, broad, and deep problem with many facets and numerous formulations, and can be addressed with a variety of objectives and constraints. The techniques, approaches and algorithms are the first step toward developing viable methodology, algorithms, and tools for addressing Location Discovery. In this chapter, we propose several research directions that deserve future investigation including characterization of mobile trajectories, design of infrastructure for characterization.

Techniques for Enabling Efficient Location Discovery

We address two problems in this chapter: (1) how to maximally improve the accuracy of location discovery (LD) in static and mobile ad hoc networks by placing one or multiple additional nodes and (2) how to create and maintain location discovery infrastructure (LDI) for static and mobile users. For this purpose, we have developed two new types of methodology. The first one coordinates the construction of statistical models and optimization techniques, such that the models facilitate rapid and efficient optimization, which leverages the properties of the models. Specifically, we enforce that all our models are piecewise linear and monotonic, and all our algorithms employ piecewise Linear Programming. The second type of methodology uses constraint manipulation to create flexible LD that supports user objectives while balancing security and privacy. We have constructed four types of models: (1) distance measurement error models, (2) LD error models, (3) indoor and outdoor environment models, and (4) individual and group mobility models. Our approach is evaluated on sets of distance measurements produced by actual deployed networks, and we analyze the performance in terms of accuracy, scalability, security, and privacy.

Beacon Positioning and Operations

We address the problem of placing a minimal number of beacons in a complex terrain in such a way that any arbitrary stationary or mobile node can locate itself within specified error and time limits. The starting point for our approach is the data-driven distance measurement, environment, and localization error models. These models are used to create the OFs for the three phases of our approach: (1) beacon placement, (2) beacon grouping for simultaneous activation, and (3) beacon scheduling. We prove that each of these three tasks is NP-complete and create heuristics and (integer) linear programming algorithms. In the first phase, we construct nonparametric statistical localization error models that capture the joint conditional probability of expected location error based on two properties: number of distance measurements and the third largest angle of all neighbors. The beacons are placed so that the location errors are minimized for a representative set of nodes. In the last two phases, we address the problem of which beacons, when, and how to broadcast acoustic signals, so that maximum number of unknown-location nodes can calculate their distance measurements as frequently as possible. We analyze the scalability of our approach and its dependency on parameters such as network connectivity and size and beacon density. Location discovery has received a great deal of research attention in wireless ad hoc community because of its role as an essential enabler required by other tasks such as routing and data fusion. A number of exceptionally creative and effective LD approaches have been demonstrated. A closely related problem is building and operating permanent or ad hoc Location Discovery Infrastructure (LDI), where the goal is to place a small number of beacons in such a way that any other node at an arbitrary location can accurately calculate its location. Surprisingly, this problem received rather little attention regardless of its apparent usefulness and technically challenging structure. The problem is challenging because it consists of three-layered NP-complete sub-problems and the statistical uncertainty of the distance measurements.

Location Discovery in Presence of Insufficient Number of Measurements

There are several reasons why offline models are important. First, they enable us to learn about the properties of the error distribution functions, which can be used for faster and less expensive online model development. Second, in many actual cases when we have beacons, whose exact locations are provided by GPS devices, the distances between the beacons can be measured. Consequently, we can easily construct an online model based on the measurements among beacons using the same approaches. Finally, we show how one can iteratively deduce error models by interleaving Location Discovery and error modeling. We quantitatively compare the impact of location accuracy based on offline and online error models.