Grantee Poster Presentations

Overview | Agenda | Attendees | Presentations | Breakout Session Summaries | Grantee Poster Presentations

Poster Abstracts

Back to top

Generating Virtual Random Walkthroughs Of A Remote Scene From A Network Of Cameras

Narendra Ahuja and Yoshihisa Shinagawa
University of Illinois at Urbana-Champaign
Beckman Institute
Urbana, Illinois 61801
n-ahuja@uiuc.edu
sinagawa@uiuc.edu

This project is aimed at producing novel images of a 3D scene from arbitrary new viewpoints using a network of strategically placed cameras. The cameras provide a sparse set of compressed panoramic snapshots, or sample images, of the scene. A major application of the proposed work is to enable walkthroughs of the scene by generating the images of the scene along a trajectory dynamically chosen by a remote user. The focus of the proposed work is on the acquisition of panoramic sample images, their compression, and their interpolation (or extrapolation) for producing virtual images of the scene from arbitrary new viewpoints. Development, implementation and addition of a stereo panoramic camera is also planned. The amount of information on which the walkthroughs are based, and therefore the amount of information to be transmitted from the scene to the location of the virtual environment, is desired to be as small as possible. Nearby cameras in the camera network provide images that are correlated and therefore partially redundant, thus providing an opportunity for joint compression and efficient transmission. The known topology of the camera network allows the determination of which cameras are to be used to generate the walkthrough image at any given point along the user drawn trajectory. The desired intermediate view of a single object is estimated from 2D images but the estimate is modified using information about mutual and self-occlusion among scene surfaces. The use of the 2D images significantly reduces the computation performed compared to other approaches. Applications of the proposed work include video surveillance, virtual museums and video teleconferencing.

Back to top

Design Steps For Prototyping Smart SAW Sensors And A Burst Transceiver For Interrogation

Atashbar M. Z., Bazuin B. J., Krishnamurthy S.
Department of Electrical and Computer Engineering
Western Michigan University

Sensors and sensor systems are used in a wide range of industrial and consumer applications, providing real-time information that allows decisions to be made when and where required. Sensors allow industrial processes and applications to be more cost effective, reliable, and safe. In many applications, sensors and data collection processing must be distributed and placed in inhospitable or inaccessible environments. This complicates sensor and system design by requiring devices with small size, rugged construction, an efficient power supply for active components, communications for derived information, and simple installation with a minimum amount of infrastructure and overhead.
Wireless surface acoustic wave (SAW) sensor systems composed of distributed wireless SAW sensors, sensor communication transceivers, and a centralized host processor can meet the design goals and challenges described. New generations of SAW microsensors promises to provide measurements of a wide range of physical and chemical parameters, including temperature, pressure, chemical concentration, gas concentrations, etc. Compatible with integrated circuit processing techniques, SAW devices can and will be combined with active circuitry into integrated smart SAW sensors.
This poster describes a prototype transceiver that is used to provide power, interrogate, and receive responses for a range of SAW sensors together with the design and frequency response simulation of SAW devices using MATLABTM. The general design requirements for a wireless sensor transceiver will be defined and the useful operating range based on the signal power levels will be discussed. The system architecture capable of transmitting a burst of RF energy, receiving the RF sensor response, and then quadrature downconverting, digitizing, and collecting the data for processing will then be described. For the design and simulation of SAW devices, two different approaches namely the equivalent circuit approach and the transmission matrix approach will be described for the three SAW configurations namely delay line, one-port and two-port resonators along with the simulated results.

Back to top

Robust Data Dissemination In Wireless Sensor Networks

Prof. Saurabh Bagchi, Gunjan Khanna, Yu-Sung Wu, Yen-Shiang Shue,
Jen-Yeu Chen, Hakeem Ogunleye, Issa Khalil
Purdue University
Electrical and Computer Engineering Department

A large part of the job of wireless sensor networks is to convey data, primarily from the sensor nodes to cluster heads or base stations, or in the reverse direction. For several classes of deployments (environmental safety monitoring, seismic monitoring, etc.), the reliability of the data dissemination process is critical. The wireless sensor environment, however, is inherently error-prone due to node and link failures, which may be permanent or transient. The environment is also often hostile and the network is subjected to malicious attacks, such as jamming or eavesdropping. In this poster, we demonstrate our work in reliable and secure data dissemination. There are three complementary approaches that are used. In the first approach, a protocol called SPMS (Shortest Path Minded SPIN) transmits the data after exchange of meta-data information to determine which nodes are interested in the data. The data dissemination proceeds in multiple hops using the lowest possible transmission energy at each hop. In the second approach called SensorNMR, the natural redundancy that is present in sensor networks due to the high density of the sensor nodes and the broadcast nature of wireless transmission is used to form multiple redundant paths to transmit data to the base station or cluster heads. Voting and error masking are done at the end point and innnovative techniques are used to decrease the energy consumption. The third approach enables protecting the messages through cryptographic encryption and authentication by making it feasible to do symmetric key cryptography in sensor networks by providing a scalable, energy efficient and secure method of key distribution.

Back to top

Fast, Low-power Image Compression Techniques and Architectures for Mobile, Wireless Agents

Dr. Amy Bell Dr. Joan Carletta
Virginia Tech University of Akron

The JPEG2000 standard provides significantly better performance than its predecessor, JPEG, for the compression of digital images; algorithmically speaking, JPEG2000 is the ideal choice for real-time image transmission from mobile, wireless sensors. However, the effective implementation of the sophisticated JPEG2000 components—like the discrete wavelet transform (DWT) and EBCOT quantizer—presents both hardware and signal processing challenges. In this presentation, we illustrate our new filter design techniques and hardware architectures for fast, small, high quality implementations of JPEG2000’s DWT component on field programmable gate arrays (FPGAs). This work represents the first step in realizing the future ultimate vision: a wireless, mobile, lightweight, low power, low cost, easily updated, handheld/wearable/embedded device that quickly transmits and receives high quality image and video data with other devices in a wireless network.

The performance of a JPEG2000 hardware codec depends on the precision with which the DWT filter coefficients are approximated in fixed-point representations. Greater precision implies better compression performance, but at the cost of larger, slower hardware that consumes considerable power. We present a new algorithm—called “zero compensation”—for quantizing (i.e. approximating in fixed-point) the filter coefficients in a “convolution” approach to computing the DWT. This method attempts to preserve the DWT filter bank properties; we show that it enables the design of high-performance image compression filter banks with small, fast hardware. Next, we present a new polyphase filter bank structure that doubles the throughput while maintaining the high quality compression performance of zero compensation. However, this higher throughput polyphase structure requires more hardware and power than the non-polyphase structure.

Finally, we compare our best polyphase, convolution structure with the “lifting” approach to computing the DWT. Lifting is also a polyphase scheme and so it offers the same, doubled throughput of our polyphase, convolution implementation. The inherent orthogonality of the lifting structure implies that it has the unique advantage of synthesis inverting analysis in the DWT even after filter coefficient quantization. However, this orthogonal structure cannot exploit our new zero compensation technique. We present several methods for designing optimal lifting filter coefficients and compare their compression and hardware performance with our best polyphase, convolution technique. Our results indicate that the lifting implementation provides the best compression performance and requires the least power; however, the polyphase, convolution architecture exhibits the highest throughput and smallest hardware.

Back to top

Motes with Intelligent Antennae

William J. Chappel: chappell@ecn.purdue.edu and Chin-Lung YANG: cyang@purdue.edu
Department of Electrical and Computer Engineering
Purdue University

Sensor networking envisions a promising future in which scattered smart dust sensors combine to form low power, adaptive, dynamic, and self-configurable wireless networks capable of relaying their data and monitoring their external environments wirelessly. Since the wireless channel has numerous time- and location- dependent variables, such as environmental variability, node variability, traffic variability, and attacker induced variability, a very effective way for the nodes to cope with these variability is to equip them with diversity schemes which attempt to alter the nature of the channel by changing the nodes' electromagnetic characteristics.

In the RF portion of the sensor, a full range of diversity, including polarization, spatial, directionality, and node mobility, will be utilized to create electromagnetic diversity in order to maximize the channel path and the communication capability between nodes. The system benefits from each method will be analyzed to produce the smallest, most efficient diverse RF front end possible. However, this electromagnetic diverse RF front end is difficult to create due to the limited options provided by the electrically small antenna design. In our research, we propose to provide in-built support in the nodes to tolerate the uncertainty of different antenna dimensions. More importantly, we'll focus on the design of intelligent antenna by employing unique techniques to minimized form factor for a given scanning capability, to provide a high gain antenna for low power operation, to build up a narrow beam pattern to reduce the lower interference, and to create diversified antenna to be able to adapt to environment and networking needs. Equipped with this sort of intelligent antenna, our proposed sensor node will be capable of not only tolerating uncertainty, but also exploiting the uncertainty to its advantage.

As it is well known, the efficiency of traditional electrically small antennas, such as short dipoles or loops, is significantly reduced due to near field energy storage. High-K dielectric antennas can overcome this problem. But few low-loss high-k materials are available in nature and fabricating unique three-dimensional structures with them is difficult. The use of effective dielectric substrate can create novel electrically small antenna with good efficiency. Effective dielectric constant in a confined region can be obtained by local control of the periodicity of a dielectric material through advanced ceramic manipulation. With this method, artificial composite material with desired properties can be produced to replace the material found in nature. The low efficiency of electrically-small high-k dielectric antenna caused by surface waves can be easily overcome by the use of the patterned periodic substrate in which only the substrate area beneath the antenna has a high dielectric constant. Furthermore, the use of effective dielectric constant materials on antenna design will be combined with low loss switching mechanisms (ex. MEMS switches). This combination will create electrically small radiators which can adapt to the unique needs of the sensor network and provide high efficiency communication links as well.

The end goal of our proposal is to create intelligent RF front ends focused on adaptive antennas that will give realistic radiation parameters for use in optimization at higher levels in the network stack. The proposed sensor network has the power of superior re-configurability that allows for uncertainty management and power optimization. And only the size and the cost of the node itself will be the ultimate limit for this next generation sensor network.

Cooperative Control for Sensor Network Organization

Ryan Abel, University of Iowa
Soura Dasgupta, University of Iowa
Bob Bitmead, University of California, San Diego
Jon Kuhl, University of Iowa
Tamer Basar, University of Illinois

A central task of every intelligent, self organizing sensor network is to achieve a topology, defined by its geometry and the manner in which the constituents exchange information, that permits it to achieve its objectives with optimal facility. We view the attainment of such an optimal organization as being fundamentally a cooperative control task.

Thus, this proposal focuses on the organization of large, mobile, ad hoc reconfigurable networks for coordinated sensing or surveillance, using a coordinated control approach. The archetypal applications problem which serves as a benchmark is the coordinated control of a number of sensor-bearing vehicles, such as a fleet of autonomous vehicles that performs cooperative. We assume each unit must exchange information through peer to peer, multihop wireless communications to perform the requisite tasks in a potentially unfriendly environment.

Issues of fault tolerance, network architecture, collision avoidance and the impact of information quality on control algorithms, are discussed using the glue of Model Predictive Control.

Multifunctional Wireless Microsystems with Assembled Nanosensors

A. Narayanan1, N. DiLello2, V. Deshpande2, M. Riegelman3,
H.H. Bau3, S. Evoy2, S. Raman1

1. Bradley Department of Electrical and Computer Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA 24061.
2. Department of Electrical and Systems Engineering, The University of Pennsylvania, Philadelphia, PA 19104.
3. Department of Mechanical Engineering and Applied Mechanics, The University of Pennsylvania, Philadelphia, PA 19104.

Wireless Sensor Microsystems are expected to revolutionize the security, control and monitoring operations of commercial, military and industrial systems. We present a technique that enables integration of nano-scale sensor devices with low-power CMOS Integrated Circuits, in particular, sensor readout, data conversion and wireless communication functionalities. At present, we have demonstrated the assembly of nanowires such as metallic nanowires and carbon nanotubes onto lithographically defined electrodes fabricated on a Si chip for proof of concept purposes. The nanowires are assembled between two electrodes by applying an alternating electric field across the electrodes. The placement and alignment of the nanowires are defined by dielectrophoretic forces that act on the nanowires. Characterization of the assembled wires and carbon nanostructures is also presented. We plan to extend this work to provide a new dimension in multifunctional sensor systems. The technique described above allows for the sequential assembly of diverse functionalized nanosensor elements onto a single-chip system. For example, the sensing nanostructures can be straightforwardly integrated with a low-power frequency-hopped spread spectrum transmitter. The proposed transmitter architecture allows a large number of sensor nodes to communicate within each other and also with the central control unit at low-to moderate data rates. With integration of diverse functionalized nano-scale sensors with low-power read-out and data communication system, a versatile and commercially viable low-power wireless sensor system can be realized.

A Simulation-Based Test Bed for Networked Sensors in Surface Transportation Systems

Richard Fujimoto1 and Randall Guensler2
1College of Computing, 2School of Civil and Environmental Engineering
Georgia Institute of Technology

Urban transportation systems face important challenges concerning safety, delays due to congestion, pollution from vehicle emissions, and vulnerability to natural and man-made disasters. Sophisticated, integrated sensor/actuator systems offer the potential to make significant inroads in alleviating these problems. Vehicle-to-vehicle communication via ad hoc wireless networks present new opportunities for innovative applications with reduced public infrastructure investments. This project is exploring the realization of simulation-based test beds for rapid evaluation and integration of sensor and actuator systems in Intelligent Transportation Systems (ITS), and use this test bed to explore architectural alternatives for data networks supporting ITS applications. An integrated distributed simulation approach is being utilized, focusing on interoperable simulations of transportation infrastructures, wired and wireless communication networks, and distributed computing applications. A large-scale application of the CORSIM vehicle simulation model provides relevant vehicle speed and position information for scenario analysis. The project also incorporates data from a pool of more than 500 instrumented vehicles operating in the Atlanta metropolitan area, which provide actual second-by-second data on vehicle speed, acceleration, position, and engine operating conditions. Data from road sensors in the Atlanta area will also be used for model and scenario development and validation of simulations.

Generic Autonomous Platform for Sensor Systems (GAP4S): An Environmental Monitoring Application

Franco Maloberti, Andrea Fumagalli, Jin Liu, Murat Torlak, Stefano Gregori, Marco Tacca, Devrim Aksin, Balkan Kecicioglu, Paolo Monti, Tianhon
Erik Jonsson School of Engineering and Computer Science
University of Texas at Dallas

Networks of wireless integrated sensors are used to estimate distributed parameters related to a variety of applications, such as security, medical monitoring, installation diagnostics, chemical and biological detection. A number of recent research efforts opened the way to relatively inexpensive wireless micro-sensors. Some solutions rely on limited-life batteries to provide long-term sensor and network functionality. Other solutions envision short transmission range sensors (few meters) harvesting the necessary energy from various sources (solar, vibrations, acoustic noise).

The Generic Autonomous Platform for Sensors (GAP4S) project explores a novel approach for wireless sensors that is complementary to the above studies. This project provides multiple integrated sensing functions, medium transmission range sensors (tens of meters), remote re-charge of the sensor micro-battery, and end-to-end reliable access. The above objectives are obtained by combining: 1) a single micro-board with full-custom ultra low-power IC's (sensing devices and radio frequency (RF) transmitter), high efficiency microwave receiver, micro-battery, and signal processing capabilities, 2) a unique use of the microwave signal to simultaneously re-charge the sensor micro-battery, synchronize data slots, and provide the return link from the base station, 3) a power-aware and reliable protocol stack spanning across both the wireless sensor network and the existing network infrastructure (the Internet).

A full-custom chip was designed and fabricated using 0.18 um CMOS technology on the micro-board. A combined optical sensor (a low-resolution digital camera together with an analog motion detector and a low power data converter) and an RF transmitter (power amplifier and voltage controlled oscillator (VCO)) are on the chip. The data converter, used both for recharging the pixels and for 7-bit data conversion, has 10 uW power consumption. The LC-tank VCO has 150 uW power dissipation and a 100 dBc/Hz phase noise (@1 MHz offset). The class E power-amplifier has power efficiency higher than 80%.

A 433 MHz FSK modulation has been chosen in the communication link sensor-base-station. FSK eliminates the need for a linear power-amplifier, thus reducing power consumption. The multiple-access of the sensors is accomplished with a hybrid frequency and time division scheme. An empirical path loss model for distances up to 100 meters has been developed. The smart antenna base-station has been designed.

Two tasks were addressed in the field of network architecture and protocols. First, an original approach to maximize the lifetime of a cluster-based system was examined. A greedy algorithm able to determine the optimal assignment of nodes to cluster-heads was designed. Second, the end-to-end communication reliability between the base-station and the end-user was investigated. A dynamic algorithm able to reduce the overall network blocking probability by efficiently making use of network spare resources was presented.

The key features of the GAP4S system - e.g., the possibility of using different plug-in sensors and the low maintenance guaranteed by recharging the micro-batteries with microwaves - will provide valuable instruments to practitioners to design wireless sensor networks in the years to come. Soon it will be possible to design cost effective, medium range and reliable wireless sensor networks whose applications span from buildings and airports, to monument control and industrial-agricultural activities.

Back to top

Centralized and Distributed Fault Identification in Discrete Event Systems

Christoforos Hadjicostis
Electrical and Computer Engineering
University of Illinois at Urbana-Champaign

We study fault identification in discrete event systems that are described by Petri nets. We consider faults in both Petri net transitions and places and assume that system events are unobservable but that system state is periodically observable. Our approach employs redundant Petri net embeddings that are designed in a way that enables fault detection and identification to be performed in a centralized or distributed manner using algebraic decoding techniques. We first develop an algorithm for solving a system of composite power polynomial equations and then combine this method with traditional decoding techniques for Reed-Solomon codes in order to efficiently identify multiple transition and place faults using a centralized diagnosis mechanism. More specifically, using a redundant Petri net embedding with 2k additional places, we can simultaneously identify up to 2k-1 transition faults and/or up to k place faults (that may occur at various instants during the operation of the Petri net) with worst-case complexity of O(k2m + kn) operations (where m and n are respectively the number of transitions and places in the given Petri net model). The proposed diagnosis schemes are extended to distributed settings in ways that require negligible additional hardware.

Back to top

Wireless Sensor Networks with Rayleigh Fading Channels

Often, a geometric "disk model" is used to model the link or connectivity in wireless sensor networks. While this model has led to important fundamental results, it is apparent that the wireless link does not obey such a deterministic disk abstraction. Variations in the path loss have to be taken into account even at higher layers in the protocol stack, since the stochastic nature of the wireless channel has an impact at the link, MAC, network, and transport layers. In fact, the disk model has several shortcomings that lead to unrealistic assumptions: 1) It suggests that the reception probability over a multihop connection is 100% if every receiver is in the range (disk) of the transmitter. A consequence of this is that the energy benefit of splitting a hop in two shorter hops is larger than it actually is. 2) Since interference is usually modeled using the same disk model (possibly with a bigger radius), it suggests that power scaling across the network results in a larger number of collisions. 3) It does not account for the accumulated interference of a large number of distant nodes. 4) Since there is no benefit in routing over different path or retransmissions, the "disk model" does not permit the performance assessment of spatial and temporal diversity schemes.

To overcome these shortcomings of the "disk model", we suggest a link model that is based on a Rayleigh fading channel. Under Rayleigh fading, the received power is exponentially distributed. With the standard assumption that a packet is received if the SINR is bigger than some threshold, this permits a factorization of the packet reception probability into a term that only depends on the noise and a term that only depends on the interference. As a consequence, performance analyses can be carried out separately for a zero-interference and a zero-noise network.

We discuss several important consequences of fading channels. At the MAC layer, two extreme schemes are considered, the simplest random access scheme (ALOHA-like) and the globally optimum scheme. A performance comparison shows that the maximum throughput benefit of sophisticated MAC schemes is a factor of two. For routing, we demonstrate that there are benefits of using longer hops rather than nearest neighbor routing, if both schemes are given the same delay bounds.

Back to top

Architectures for Resource Discovery, Query Resolution and Rendezvous in Large-Scale Wireless Networks: Design and Analysis Framework

Ahmed Helmy
Electrical Engineering Department
University of Southern California

This research project addresses issues of architectural design and evaluation for classes of service in large scale wireless (ad hoc and sensor) networks. The first class of service targeted in this project is that of 'small transfers' including resource discovery, query resolution and transactions. A novel, contact-based, approach is taken to introduce efficient power-aware routing protocols for small transfers. Several contact-based protocols, e.g., CARD, MARQ, TRANSFER, are designed and evaluated showing significant power and overhead savings for small transfers. The second class of service targeted is that of storage-retrieval (or data-centric storage) applications. A new architecture, called 'Rendezvous Regions', is proposed that is based on consistent geographic mapping from the 'resource' (stored or sought) to a 'region'. Extensive evaluations and comparisons for both the above architectures are carried out, to identify ranges of operating conditions under which the proposed ideas are of significant advantage. In addition, several new studies are conducted on geographic routing with inaccurate (or inconsistent) location information. Several pathological scenarios are uncovered with existing geographical routing approaches.

Finally, a framework for studying protocol behavior under a 'rich' set of mobility models is designed. Several mobility models (e.g., manhattan, freeway) are introduced, along with various metrics (e.g., spatial/temporal correlation) to capture significant characteristics of the mobility models. Studies are also conducted to characterize and understand the effects of mobility patterns on path statistics and protocol behavior. The 'IMPORTANT' mobility tool, in which our proposed models were implemented, was used in the evaluations.

Back to top

Intelligent RF And Microwave Front Ends

D. Goshi, K. Leong, Y. Wang and T. Itoh
Electrical Engineering Department
University of California, Los Angeles

We present several projects on RF front end hardware configurations that enhance signal processing capabilities, including the Signal Processing Antenna supported by NSF ITR and related projects. Under the signal processing antenna, we have developed Single RF-channel receiving antenna array with Spatial MultIplexing of Local Elements (SMILE) to reduce expensive RF front end hardware to only one channel while the inputs to antenna array elements are spatially multiplexed. It is experimentally demonstrated that this SMILE array has the same capability as the conventional array follwed by as many RF front end circuits as the number of antenna elements.

We have developed several smart antennas and experimentally demonstrated. In order to cope with the DSP congestion, we have implemented hybrid architecture in which the RF signal processing is done in an analog mode while the beam control is carried out digitally.

In the area of intelligent transmitter, we have developed an adaptive power combining scheme to increase the overall power added efficiency of the power amplifiers at the last stage of the transmitter. For a low-level input, only one amplifier is used while all four amplifiers are combined in parrallel to deal with a high-level power inputs in such a way that the amplifiers are operated at a best power added efficiency regime.

Finally, as a completely analog signal processing antenna, we present a retrodirective array based on phase conjugation. This array reflects an incoming wave back toward its source without prior without prior knowledge of the source. Since this operation is carried out completely in an analog fashion, the processing speed is much faster than a digital scheme. In addition, both uplink and downlink superposition of data signals for duplex communication is possible.

Throughout these projects, we emphasize multidisciplinary approach in which digital and analog techniques are combined in a collaborative manner to take advantage of each. Simple insertion of digital technique does not alleviate importance of analog circuits.

Back to top

Optimal Deployment And Energy Management Of Wireless Sensor Networks

Koushik Kar
Department of Electrical, Computer and Systems Engineering
Rensselaer Polytechnic Institute

With significant advancements in device technologies, small, low-powered, low-cost sensors have become feasible today. These devices can be deployed in large numbers to form a dense networked environment, with the objective of gathering data for various purposes. Therefore, even though an individual sensor device is expected to be cheap, the sheer number of the sensors would make the monetary cost of deployment significant. Hence the question of cost-optimal deployment is of prime importance. In our research, we are investigating some fundamental theoretical questions regarding optimal deployment of large-scale sensor networks, and explore how our theoretical results can be applied to develop effective deployment strategies in complex practical scenarios. More specifically, our research focuses on obtaining sensor deployment strategies that minimize the number of deployed sensors while meeting desired objectives on connectivity, coverage and network lifetime. We are investigating such problems in both deterministic and random deployment scenarios, and also in more realistic, intermediate scenarios with controlled randomness.

For energy-constrained applications, the problem of optimal sensor deployment is closely related to the problem of optimal energy management and routing. A non-optimal energy management and routing policy would result in a redundant deployment of sensors to meet the same application-specific lifetime constraints.

Our current research also focuses on developing optimal and near-optimal policies for energy management and routing in sensor networks. The energy management policy determine which nodes remain powered on at any point in time, and which nodes (amongst the ones that are powered on) are involved in sensing, the rest being involved only in routing. We are investigating these energy management and routing questions in cases where the statistical characteristics of data generation rates are known in advance, as well as in cases where no such information is available a priori.

Back to top

The Use Of Ontologies For The Self-Awareness Of Communication Nodes

Mieczyslaw M. Kokar, David Brady, Kenneth Baclawski, Jiao Wang, Leszek Lechowicz
Northeastern University
Boston, MA 02115

Although migrating algorithms from hardware to software can increase the level of functionality, it does not necessarily change the fact that the communication protocols are established at design/development time. We are investigating an approach to establishing communication by explicitly maintaining self-awareness and communication of knowledge about the operation of the communication nodes. The self-awareness and communication of knowledge is based on the maintenance of an explicit, declarative knowledge base or ontology of communication. Hence, we refer to the concept as Ontology Based Radio (OBR). In this paper we will show some preliminary results of our research.

The main idea of OBR is that the communication nodes “understand” the contents of information to be transferred, their own capabilities and capabilities of the destination units. This understanding is expressed using ontologies. Ontologies are at the core of semantic information processing. An ontology captures the basic terminology (concepts) of the domain of interest and the relationships among the concepts. Ontologies can be expressed graphically using the Unified Modeling Language, represented using the DARPA Agent Markup Language for ease of interchange, and processed either off-line using a theorem prover or in real time using an expert system engine.

The proposed approach is based on the model-driven architecture implemented by the means of ontologies, DAML-based annotations and Java’s reflection capabilities. Each software module can be queried about its structure and contents using a DAML based query. It can then reply to the queries by analyzing its own structure using Java’s reflection and the system’s inferencing capability. Since our goal is to give a proof of concept for this approach, in the first stage we are using a generic theorem prover for drawing inferences. This approach is obviously not appropriate for real-time implementation.

In this paper we will show an example of such a functionality in which two nodes exchange information and then reason about the multipath structure. The net result is that after analyzing the multipath structure nodes can improve the efficiency of communication.

Back to top

Overload Management In Sensor-Actuator Networks

Michael Lemmon
Department of Electrical Engineering
University of Notre Dame
lemmon@nd.edu

Overload management policies avoid network congestion by actively dropping packets. This paper studies the effect that such data dropouts have on the performance of spatially distributed control systems. We formally relate the spatially-distributed system's performance (as measured by the average output signal power) to the data dropout rate. This relationship is used to pose an optimization problem whose solution is a Markov chain characterizing a dropout process that maximizes control system performance subject to a specified lower bound on the dropout rate. We then use this Markov chain as a novel quality of service (QoS) constraint on the sensor-actuator network. Several management heuristics are proposed for enforcing this novel QoS constraint and preliminary evaluations of their performance is presented.

Back to top

Biomimetic Sensors for Autonomous Underwater Vehicles using MEMS Artificial Haircells

Chang Liu
Department of Mechanical and Industrial Engineering
University of Illinois, Urbana-Champaign

Water covers approximately 70% of the earth surface. However, this majority of the planet is under-explored and poorly understood. This is largely a result of the significant challenge posed by the underwater environment, with its high pressure, unpredictable currents, nearly freezing temperatures, attenuation of electromagnetic waves and light, and harsh chemical compositions. A variety of scientific, military and industrial tasks call for tools that function in the ocean environment. Autonomous Underwater Vehicles (AUVs) are important platforms for carrying out sophisticated underwater tasks. Currently some tasks are not suitable for AUVs due to technical limitations. More advanced applications demand small, integrated, power efficient and potentially networked AUVs to increase the effectiveness and reduce the costs. A key step in moving toward this vision is improved sensors for AUVs. This project aims to develop integrated underwater sensors inspired by the fish lateral line system. Prototypes of sensors will be built and used to investigate the fluid mechanical behavior of such sensors through computational and experimental means. The potential applications include the detection of hydrodynamic wakes, tracing of chemical plumes, and determination of hydrodynamic properties at the interface of fluid and structures. The proposed effort will focus on one problem - that of detection of hydrodynamic wakes.

Planned research activities consists of three integral components: (1) The development and improvement of biomimetic integrated microsensors; (2) The validation and prediction of sensor performance using full scale direct numerical simulation and reduced-model simulation; (3) The experimental validation of sensor performance and computational results. The artificial haircell sensor is a modular building block for realizing application specific integrated sensors.

Back to top

Virtual Sensor-Actuator Arrays (VSAA)

PIs: Bruce H. Krogh, José Moura, Rohit Negi
Graduate Students: Haotian Zhang, Wei Zhang
Department of Electrical and Computer Engineering
Carnegie Mellon University

The objective of this research project is to develop new computational methods for processing information on highly distributed arrays of wireless sensors and actuators. We propose to develop algorithms that transform information from the spatial and temporal domains of the data from the physical sensors into a virtual sensor-actuator array domain in which command and control computations can be performed with respect to the application-specific frames of reference. The principle motivation for the introduction of the virtual sensor-actuator array is to make it possible for domain-specific computations to be performed without regard to the details of the spatial and temporal distribution of the actual sensor-actuator array.

The poster presents the elements of our approach. We are developing methods for estimating parameters in physically-based finite dimensional models to perform estimation and prediction of spatially distributed signals. We illustrate our approach for the problem of temperature estimation. In previous work we estimated temperature distributions from data collected using Motes sensors using 2D curve-fitting techniques. This approach leads to significant errors, particularly at boundaries of the region. In our current work, we use parametric system identification to estimate parameters in a resistive-capacitive lumped parameter model of the temperature dynamics. In this model, the sensed values are the inputs and the temperatures to be estimated from the model are the outputs. Currently, data for the system identification and temperature estimation experiments are generated from a simulation model of a simple room. We are also applying factor graphs and the sum-product algorithm for intelligent sensor fusion. The poster presents simulation results for target location estimation using these techniques. Here the goal is to compute the likelihood of a given target being detected based on the soft decisions (probability distributions) of the individual sensors.

Back to top

Towards Sensate Media and Electronic Skins - Testbeds For Very High Density Sensor Networks

Joseph A. Paradiso, Joshua Lifton, Michael Broxton
Responsive Environments Group
MIT Media Laboratory

Today’s canonical sensor network generally involves scores of sensor/processor nodes communicating peer-peer across internode distances of meters or more. While this tends to serve applications ranging from battlefields to smart buildings, there are many other scenarios where a higher node density becomes interesting. A simple example is a “smart wallpaper”, with hundreds of nodes spaced within centimeters of one another – as one can readily access each node and easily control the sensor stimulus and background environment, this structure can serve as a very effective testbed for the development of distributed estimation and decentralized sensor processing algorithms.

Accordingly, we have built such a system, which we call “PushPin Computing”. As its name suggests, a pair of insulated-jacket conductive pins protruding from the rear of these nodes provides power when they are pushed into a large piece of laminated conductive wallboard, avoiding the need for batteries. The nodes communicate with their immediate neighbors via IR once inserted, and form an ad-hoc network across the wallboard. The wallboard substrate accommodates large numbers (e.g., over 100) of these pushpins, and they can be trivially reconfigured, allowing different network densities and geometries to be easily investigated. Likewise, the pushpins themselves have a configurable, layered structure, allowing the processor, communications system, and sensor suite to be easily swapped. The sensor layer mounts at the top of the pushpin, accommodating many modalities (e.g., photonic sensors, acoustic pickups, capacitive proximity detectors, temperature monitors) while also providing some limited capability to display status at each node via a small array of visible LEDs.
A dense configuration of smart sensor nodes on a wallboard begins to hint at a new kind of sensate material, which one could loosely think of as a multisensory electronic “skin”. Inspired by the analogy to biological skin, where signals from huge numbers of closely-spaced receptors are preprocessed (e.g., inhibited or enhanced) in the nervous system before reaching the brain, processors talk to their neighborhood to reject background while isolating and parameterizing sensor stimuli, then rout the results to external connections at the network’s perimeter. Such structures promise to avoid difficulties associated with scaling standard multiplexing schemes to very large systems.

As a first step, we have built the “Trible” (Tactile Reactive Interface Based on Linked Elements), which has the form factor of a soccer ball, tiled with a multimodal sensate "skin" consisting of 32 networked tiles. Each measures pressure at 3 locations, local temperature, local sound via a microphone, local illumination, and dynamic tactile stimulation with up to 12 channels of protruding, touch-sensitive piezoelectric "whiskers". Each tile can also respond with a small audio speaker, a pager vibrator, and a bright RGB LED. There is no central processor - each tile talks to its neighbors through conductors in the frame. The Trible is a research platform for the application of decentralized control and distributed estimation to human-computer interaction.
Our forthcoming testbed will achieve higher density, e.g., a sheet of material with circa 100 or more multimodal sensor/processor nodes spaced under a centimeter to one another.

Back to top

Wireless Sensor Networks

Catherine Rosenberg, Sunil Kulkarni, Vivek Mhatre, Aravind Iyer
Department of Electrical and Computer Engineering
Purdue University

We discuss some of the important future applications of wireless sensor networks, and then present some of our current work. Our current work is mainly on the network dimensioning aspects (lifetime, node density and battery energy), cross-layer interaction issues (physical layer, MAC and routing, and structural characteristics and trade-offs (clustering, aggregation and communication).

Back to top

Fair, Energy Efficient and Trust level MAC Protocols for Wireless Sensor Networks

J. Sarangapani
Embedded Systems and Networking Laboratory
Department of Electrical and Computer Engineering
The University of Missouri-Rolla (UMR)
1870 Miner Circle, Rolla, MO 65401.

Protocols and algorithms for assuring Quality of Service (QoS) in wireless sensor networks are being developed at Embedded Systems and Networking Laboratory of UMR. The QoS is met using energy efficient protocols for power control and routing, accurate yet online wireless channel state prediction scheme, and distributed fair scheduling (DFS) scheme with dynamic weight updating mechanism. Trusted levels are being added to the energy efficient routing protocol. Results indicate that our proposed distributed power control (DPC) scheme consumes least energy per packet and achieves higher spatial reuse. Moreover, the proposed DFS scheme allows fair allocation of bandwidth while maintaining minimum end-to-end delay. A novel optimal link state routing scheme (OLSR) that utilizes both the trust levels and energy efficient transmission is currently being addressed. The proposed DPC, DSR and OLSR are shown to converge mathematically and result in better performance when compared to other existing protocols for wireless sensor networks.

Back to top

Modeling, Analysis and Recognition of Dynamic Visual Processes

Gianfranco Doretto, Daniel Cremers, Paolo Favaro and Stefano Soatto
University of California, Los Angeles
Computer Science Department
{doretto, cremers, favaro, soatto}@cs.ucla.edu

This project is aimed at the development of analytical and computational tools for the study of video-based models of sequences of natural images acquired from strategically placed vision sensors. Rather than building a physical model of the scene that generates a video-sequence, the project concentrates on modeling the spatio-temporal statistics of the visual signal. This model can then be useful for purposes like video coding, surveillance, robot navigation, recognition, ecosystem monitoring, medical image analysis, object tracking, image-based rendering. These goals entail the study and development of new models, and learning procedures, as well as their coupling with statistical and variational techniques.