Many applications in sensor networks require the estimation of spatial environmental fields. We focus on the applications where the estimation is done by fitting a parametric model to the field. We study the case when the parametric model structure is unknown. Instead of assuming a particular structure, we introduce uncertainty by assuming that the spatial field is one of multiple plausible models. We then set a likelihood test for the model selection and find a spatial sampling strategy that optimizes the test. The strategy finds the locations that result in a minimum probability of error in the selection of the correct model structure. The strategy is introduced by Atkinson and Fedorov and is called T-design. We present as well the benefit over the passive (random strategy) data collection.

Mains electrical lines in buildings and houses contain a wealth of information about devices that can potentially be exploited through research. However, accessing information through the circuit breaker can be quite difficult due to dangerous high voltages and lack of infrastructure for data logging. The Power Capture Board (PowCap) system was created to solve this problem of end-to-end access. It stores high quality analog signals from a circuit breaker to an open centralized database. It is composed of a customized high power tolerant analog front-end block connected to a digitizer that is an Ettus Research software defined radio board (USRP). The USRP has the ability to collect high-resolution analog voltages with a 24-bit ADC and generates 1 Terabyte of high-resolution data a day. These rich datasets are downloadable through the web and can now be generated on a daily basis to provide signal processing opportunities that did not exist before.

As IoT and drone applications become increasing popular, there exists a demand to be able to increase communications range. In addition, with possible applications in video streaming/AR/VR, the necessity for a low-latency scheme to reduce overall transmission delay becomes more critical. Therefore, we approach this problem on two fronts – one on the PHY layer to solve the long-range issue, and the other on the MAC layer to solve the latency issue. There are preexisting works on long range communications, but these use directional antennas, high powered amplifiers, and protocol customizations on the MAC layer. Unfortunately, these approaches are not suited for the spread out and mobile nature of IoT and drone applications. Therefore, we focus on the PHY layer, specifically on the concept of predistortion, to offer our solution: increased linear power that would offer a boost in omnidirectional range without the need of directional antennas. Although much work has been done to reduce overall latency, the methods do not completely address our problem. To meet millisecond level requirements for delay sensitive applications, we recognize that the long tail of the packet delay distribution poses a significant issue. Although different mitigation techniques can be developed on various layers, we choose a path that would offer maximum marketability: a software-based solution at the MAC layer that offers the maximum amount of delay reduction without changing components at the physical level. We can therefore couple our long-range solution in PHY with our low latency resolution in MAC in the form of an API that would allow for easy installation and customizations of commodity Wi-Fi systems.

In this poster, we consider two major themes in the design of sensor networks: data integrity, and sampling strategies. For the data integrity problem, we propose a signature-based fault detection system for identifying both intermittent faults and persistent faults. Data-dependent features using temporal, spatial, and spatio-temporal information that are useful for detecting faults are identified. These features are combined into signatures that characterize each of the different fault types. We also discuss the problem of simultaneous parameter estimation and fault detection. In this case, parameters must be estimated from a distribution that is truncated in various ways as a result of the fault detection algorithm, which can lead to biased estimates. We propose several methods to account for the bias in parameter estimates. For the sampling problem, we describe two on-going projects. The first one deals with situations where sampling as you move (using sampling paths) is more effective than contemplating sampling points. A PAR sensor riding on a NIMS 3D node is one such situation. This configuration is especially well-suited for sampling phenomena that exhibit latent geometric structure, such as light fields in forest understories. We will consider the case where the phenomena can be approximated by a piecewise-constant field and suggest a novel estimation approach when we have sample paths as observations. The second project considers the problem of finding a sampling strategy to optimize the selection of the correct regression model from a set of competing regression models. The solution is driven by minimizing the probability of error in the selection and consists of a sequential algorithm that directs the collection of measurements. We develop an adaptive sampling algorithm to sample the field with a set of static sensors and one mobile sensor. The algorithm aims to jointly minimize the probability of error in the selection and the mobility cost. The algorithm presented provides a significant improvement in the probability of error in the selection of the correct model over the random collection of measurements.

Little of the work in sensor network studies related to data quality has presented a detailed study of sensor faults and fault models. We provide a comprehensive look at sensor network data fault types and a unified basis for describing sensor faults backed up by real world deployment examples. We also identify several considerations one must take into account when developing a fault detection or diagnosis system. We suggest a broad framework of important considerations when developing a data fault detection system and discuss some assumptions that can be made in this context. Based upon experience and previous work we define a series of features to consider when modeling sensor data for either fault detection or fault correction. We define three main headings of types of features and explore the effect that each type has on a sensor. Then we list all common faults that we have observed in actual sensor network deployments. We provide an example of how one may use such a taxonomy of faults.