G. Pottie's research while affiliated with University of California, Los Angeles and other places

Environment reconstruction through sampling is a difficult task and usually requires a large amount of resources. In this paper, a sampling technique is presented that approaches exhaustive sampling performance with only sparse samples. The goal is achieved by combining information from sensors of different types and resolutions. Image processing techniques are employed to extract global information. This information is passed on to the local sensors to optimize the number and locations of low-level sampling points. The sampled values are then applied back to the image to reconstruct the whole field. The technique is tested in the lab setup and shown to achieve a better result than traditional sampling methods

In a sensor network, the data collected by different sensors are often correlated because they are observations of related phenomena. This property has prompted many researchers to propose data centric routing to reduce the communication cost. In this paper, we design heuristic algorithms for combined routing and source coding with explicit side information. We build a data rate model upon the observation that in many physical situations the side information that provides the most coding gain comes from a small number of nearby sensors. Based on this model, we formulate a problem to determine the optimal routes for transmitting data to the fusion center. The overall optimization is NP hard because it has minimum Steiner tree as a sub-problem. We then propose a heuristic algorithm that is inspired by balanced trees that have small total weights and reasonable distance from each sensor to the fusion center. The average performance of the algorithm is analyzed and compared to other routing methods through simulations

We implement a two-stage DPCM coding scheme for wireless sensor networks. The scheme consists of temporal and spatial stages that compress data by making predictions based on samples from the past and helping sensors. It continuously monitors the additional gain provided by samples from other sensors, and therefore can be combined with data-centric routing algorithms for joint compression/routing optimization. Backward ε-NLMS adaptation is used to better track changing environments and avoid coefficient transmissions. Several simulations are conducted to demonstrate the effectiveness of this coding scheme.

This paper studies the problem of combining tree routing and data compression with explicit side information in wireless sensor networks. We first present our network flow and data rate model based on the observation that in many practical situations, side information providing the most coding gain comes from a few nearby sensors. An optimization problem is then formulated and shown to be NP hard. It is subsequently cast as a mixed integer program. For our particular model, we examine several approximation algorithms, and compare their performances through simulations. Improvement over shortest path trees, which completely ignore the source correlation, is achieved only by judiciously merging flows of correlated data based on the coding gain information

We consider spatial fidelity in sensor networks and show that many problems in such networks are not well defined without it. The sensor networks gathering data on multiple point sources, or a distributed source such as bandlimited or nonbandlimited field are considered. The field could either be deterministic or random. For these networks, we derive the conditions leading to cooperation between the power constrained sensor nodes for the data fusion purposes and analyze the sensor density given spatial fidelity constraints and node cooperation. We also propose heuristics for sampling density for distributed sensor networks.

We address the problem of cooperative coding among pairs of transmitters and corresponding pairs of receivers. This may be used, for example, to overcome gaps in a multihop networks. We derive upper bounds on the achievable rates for several cooperation scenarios. We show that transmitter cooperation provides significantly more improvement in rate than receiver cooperation. Based on the theoretical results, we also discuss the possible practical implementation techniques.

We consider the network information theoretic problem of finding the rate distortion bound when multiple correlated Gaussian sources are present. One of these is the source of interest but some side information from other sources is also transmitted to help reduce the distortion in the reproduction of the first source. The other sources are treated as helpers and are also coded. Special cases of this problem have been solved before, such as when the reproduction is lossless, when the sources are conditionally independent given one of them, or when the number of helpers is limited to one. We consider a generalized version and show that the previously derived expressions fall out as special cases of our bound. Our results can be directly utilized by designers to choose not only how many of the available sources should actually be communicated but also which sources have the highest potential to reduce the distortion.

Networked sensors (ENS) provide a distributed monitoring approach for defense systems situational awareness, machine condition based maintenance, health care, transportation, and other applications. The ENS architecture is based on compact, intelligent, networked low-power sensor nodes. Wireless networking enables rapid distribution of sensor nodes in diverse environments. For the general applications considered here, it is energy usage drawn from fixed energy sources that limit sensor node lifetime, and since wireless network interface operations dominate energy usage, it is essential to reduce the demand for high energy wireless data transport. Thus, identification of events and the evaluation of the utility for event data transmission must be performed locally. This, in turn, requires that methods be developed for local signal processing and event detection at the node. The signal search engine (SSE) method reported here, has been developed to enable identification of target type at the site of wireless sensor nodes using acoustic and seismic signal sources and algorithms that are compatible with low power embedded systems. The SSE is "trained" with and relies on data directly collected from the field. The SSE operates with both time domain template matching and wavelet methods for target identification. These methods are evaluated and compared here in the application to data collected in the field. Results obtained from the time domain signal classification scheme and the "wavelet" method give error rates of less than 15% with minimal signal preprocessing.

We propose an entropy-based sensor selection heuristic for localization. Given 1) a prior probability distribution of the target location, and 2) the locations and the sensing models of a set of candidate sensors for selection, the heuristic selects an informative sensor such that the fusion of the selected sensor observation with the prior target location distribution would yield on average the greatest or nearly the greatest reduction in the entropy of the target location distribution. The heuristic greedily selects one sensor in each step without retrieving any actual sensor observations. The heuristic is also computationally much simpler than the mutual-information-based approaches. The effectiveness of the heuristic is evaluated using localization simulations in which Gaussian sensing models are assumed for simplicity. The heuristic is more effective when the optimal candidate sensor is more informative.

For mobile ad-hoc networks, nodes must balance a variety of tasks including sensing and communications relays. Mobile nodes might thus change location or trajectory for sensing purposes, subject to constraints on disruption of network QoS. This paper maximizes the non-communication application QoS (node motion to facilitate sensing) with communication QoS constraints (packet delay, etc.). It also gives the formulation for maximizing the throughput for the newly formed links at the new position of the node, taking link capacities into consideration. Each link is shared by multiple streams of traffic from different QoS classes, and each stream traverses many links. Although these formulations are non-linear, they can be posed as geometric programs, which can be solved efficiently. We also propose a heuristic to implement the above algorithm in ad hoc networks.