GaS : Genetic Algorithm Synthesis

GaS : Genetic Algorithm Synthesis

University of Ottawa Sensor Networks: Research Challenges in Practical Implementations, Physical Characteristics and Applications Rami Abielmona ([email protected]) Ph.D. Candidate Sensing and Modeling Research Lab (SMRLab) November 26, 2003 IEEE Ottawa ComSoc / OWRA Wireless Seminar Series Location: Auditorium @ CRC (3701 Carling Avenue) School of Information Technology and Engineering (SITE) What are sensor networks (SNETs) ? Brief history of SNETs SNET taxonomies Hardware realizations Transmission media Power consumption valuations Communication architectures SNODE physical characteristics Sensor types and characteristics Processor support Operating system support Wireless technologies SNET practical implementations SensIT Program (DARPA research) Smart Dust Project (UC Berkeley research) Intelligent Sensor Agents (SMRLab research)

Research directions Market trends / potential applications Conclusions Rami Abielmona Outline Sensor networks (SNETs) are composed of multiple interconnected and distributed sensors that collect information on areas or objects of interest Sensor nodes (SNODEs) make up each sensor network and consist of three major components: Parameter, event and object sensing Data processing and classification Data communications SNETs can be applied to a myriad of areas: Military (e.g. object tracking) Health (e.g. vital sign monitoring) Environment (e.g. natural habitat analysis) Home (e.g. motion detection) Manufacturing (e.g. assembly line fault-detection) Entertainment (e.g. virtual gaming) Digital lifestyle (e.g. parking spot tracking) Rami Abielmona What are sensor networks ? Large number of sensors

Fault-tolerance and scalability are major design factors Clustering is a potential solution to the complexity issue Low energy utilization Power-aware protocols and algorithms are being researched Could use energy-scavengers such as solar cells Network self-organization and discovery SNODEs have a high turnover ratio but the SNET does not Each SNODE needs to know its absolute, or at least relative, position, as well as its neighbours locations Collaborative signal processing Data fusion is utilized to detect, track and/or classify objects of interest (information processing) Tasking and querying abilities Data-centric vs. address-centric techniques Data aggregation and dissemination Aggregation involves transforming data to information, while dissemination involves acquiring data from the SNODEs Rami Abielmona Design for what ? SNETs MANETs SNODEs may not have a global identification Each node has a global identification Mainly utilizes broadcast communications Mainly utilizes point-to-point communications Number of nodes is relatively high Number of nodes is relatively low Limited in power, computational capacity and memory Unlimited in power, computational capacity and memory

Topology changes frequently Topology is dynamic Low-level radio frequency communications (AM/FM) Bluetooth, 802.11 and ultrawideband (UWB) Flooding and gossiping communication protocols TCP (UDP) / IP communication protocols Table 1. SNETs vs. MANETs Rami Abielmona Sensor networks vs. mobile ad-hoc networks (MANETs) Figure 1. SNET chronology Rami Abielmona SNET Chronology (1) Sound Surveillance System (SOSUS) [1] US military initiative in 1950s System of acoustic sensors at the bottom of the ocean used to detect quiet Soviet submarines Distributed Sensor Network (DSN) [2] DARPA research project in1980 Built on top of acoustic sensors with a

Resource-sharing network communication Processing techniques and algorithms Distributed software Cooperative Engagement Capability (CEC) [3,4] DARPA research project in 1995 Summoned the network-centric warfare era, where the sensors belong to shooters rather than weapons (platform-centric warfare) Goal was to provide a common operating picture imperative for distributed military operations Military Sensor Networks (FDS, ADS, JCTN, ) [5] FDS Fixed Distributed System ADS Advanced Deployable System JCTN Joint Composite Tracking Network Sensor Information Technology (SensIT) [6] Antisubmarine warfare Integrated air picture DARPA research program started in 1999 Developed new networking techniques that could be used in hostile environments Developed networked information processing (extract information from SNET data) Rami Abielmona SNET Chronology (2) Smart Dust Project [7]

UC Berkeley research project in 1999 Main goal is miniaturization Sensing and communication co-exist in a cubic millimeter package Sub-goals include integration and energy management Has spawned off many different projects including TinyOS and the Intel Mote projects AMPS Project [8] MIT research initiative in 1999 Objective is the signal and power conditioning, filtering and communication Less emphasis is placed on the sensing unit (black box) Completed in 2002 and spawned into AMPS-II (SoC package) Intel Mote Project [9] Intel Research initiative in 2000 Builds upon the Smart Dust project Attempt to build a universal embedded node platform for SNETs Smart-Its Project [10] ETH Zurich research project started in 2001 Analogy is made to Post-It notes, but using radio tags Will attach to everyday items to give them new interaction patterns and behaviors Habitat Monitoring Project [11]

Intel Research Laboratory at Berkeley collaboration started in 2002 SNODEs are burrowed under the ground and form a wireless SNET Used to non-intrusively monitor the natural habitat of sensitive wildlife (e.g. seabirds) Rami Abielmona SNET Chronology (3) Hardware realization Three main components re-appear Sensing unit Processing unit Communications unit SNODE must Consume very little power Be autonomous and low-cost Adapt easily to the environment Fit into small packaging Figure 2. SNODE internal components Adapted from [5] Todays system-on-chip (SoC) packages allow for integrated functionalities to reside on the same chip (e.g. rfPIC)

RF transceiver Data rates are very low Packets are very small Frequency re-use is very high Processor and core memory Small and fast processors ROM and RAM cores Small-footprint RTOS (e.g. TinyOS) Rami Abielmona SNET Taxonomies (1) Transmission media Radio, infrared or optical media are viable IR forces the SNODEs to have line-of-sight (LOS) capabilities which are very inefficient in SNETs Optical media forces the SNET to be interconnected using an optical fibre, resulting in an obtrusive invasion upon the environment Radio frequency (RF) media is the most suitable Standards are becoming available worldwide

Freely licensable bands (i.e. ISM) Transceivers are becoming smaller in size, cheaper in cost and lower in power consumption RF cores can be built right onto the processing unit! Rami Abielmona SNET Taxonomies (2) Power consumption valuations Sensing unit power factors Processing unit power factors [5] PP = CVdd2f + VddI0eVdd/nVT Due to leakage current Energy cost of transmitting 1 KB a distance of 100 m is approximately the same as that for executing 3 million instructions by a 100 MIPS processor!! Power saving techniques include dynamic voltage scaling, operating frequency reductions and smaller transistors (hence lower capacitance) Communications unit power factors [5] Depends on the application (e.g. temperature sensing will consume less power than visual object detection) Could be lowered by turning off the sensing unit whenever possible Pc = NT[PT(Ton + Tst) + Pout(Ton)] + NR[PR(Ron + Rst)] Start-up time (Tst) is non-negligible for RF transceivers, thus it is

inefficient to turn the latter on and off Main static power consumption parameter of the SNODE Novel techniques have to balance computation and communication Rami Abielmona SNET Taxonomies (3) Communication architectures Left figure indicates a hierarchical (military-style) communication scheme Right figure indicates a peer-to-peer scheme Figure 3. SNET sample comm. architecture Reproduced from [13] Figure 4. SNET sample comm. architecture Reproduced from [5] Rami Abielmona SNET Taxonomies (4) Communication architectures In either scheme, each SNODE is capable of collecting data, locally processing it and sending it to its neighbors/commanders A protocol stack is present on each SNODE (will be discussed in more detail in Dr. Stojmenovics presentation) Application layer Transport layer

Involves routing the data amongst the SNODEs and out the SNET Data link layer Aids in data flow control throughout the SNET Network layer Depends on the overall task being accomplished Ensures reliable communication connections between SNODEs Physical layer Encapsulates the modulation, transmission and reception of data Rami Abielmona SNET Taxonomies (5) Sensor types and characteristics [14] What are some of the sensors that could be used in the field ? Tactile and proximity Tactile feelers, tactile bumpers or distributed surface areas Capacitive, ultrasonic, microwave or optical proximity sensors Acoustical energy Sonar (sound navigation and ranging) sensors utilize the speed of

propagation of sound waves traveling through the medium to calculate the time of flight from the sensor to the object of interest Main advantages are Very low cost and easy to interface to Fairly wide dispersion angle increases probability of detection Lambertian surfaces provide excellent reflection regardless of color Main drawbacks are Attenuated by atmospheric conditions Target reflectivity is not always ideal Disturbed by air turbulence and environment temperature Rami Abielmona SNODE Physical Characteristics (1) Optical (electromagnetic) energy Optical energy sensors (i.e. infrared and laser-based systems) Main advantages are Main drawbacks are

Atmospheric absorption and scattering Environment temperature greatly affects power output of LEDs Index of refraction is a surface property of the object (i.e. variable) Magnetic compasses and gyroscopes Increased range of operation (due to narrow and collineated beam) Reduced noise and interference Fewer multipath problems Former measures vehicle heading according to true north Latter measures vehicle orientation by maintaining its balance GPS Sensor employs TOF satellite-based trilateration in order to recover its 3-D position It utilizes 4 different Earth-orbiting satellites in order to recover its absolute latitude, longitude, elevation and time synchronization Rami Abielmona SNODE Physical Characteristics (2) Environment sensors What else can we measure about our environment ?

Temperature Light intensity Smoke Humidity Pressure Acoustical noise Motion Imaging/vision Perspiration Liquid levels Weight/mass Radiation Short-term: smell, taste and time Long-term: fear, hunger, anger, happiness, sadness and beauty And what about knowledge, humour, innovation and intelligence ? Rami Abielmona SNODE Physical Characteristics (3) Processor support The following characterize our ideal processor Relatively fast execution times Low power consumption and production cost Small area footprint

On-chip memory (cache, ROM and/or RAM cores) Abundance of I/O capabilities Standard interfaces (serial, parallel, USB, ) Robust instruction set architecture Availability of development tools Testable and reliable Industrial and academic support! It is imperative to remember that this is a physical system that employs computer control for a specific purpose and not for general-purpose computation (i.e. embedded system) Rami Abielmona SNODE Physical Characteristics (4) Processo r Family Intel 8051 C Motorola 68HC11 C Motorola ColdFire C Motorola PowerPC C ARM C Atmel C Microchi p C Processor architecture RISC CISC

CISC (MCF5407) RISC (MPC5500) RISC (ARM7) RISC (AVR) RISC (PIC) Processor speed 12 MHz 16 MHz 4 MHz 33 MHz - 333 MHz Up to 300 MHz 50 MHz Up to 16 MHz Up to 40 MHz ROM size 4 KB 8 KB 12 KB 16 KB ICache, 8 KB DCache 4 MB Flash

40 KB 192 KB Flash Up to 128 KB Flash Up to 512 bytes RAM size 128 bytes 256 bytes 512 bytes 4 KB SRAM 128 KB 4 KB SRAM 4 KB SRAM Up to 368 bytes I/O capabilities 4 8-bit ports 5 8-bit ports 16-bit ports N/A Up to 75 GPIO Up to 53 GPIO Up to 33 GPIO

Interfaces UART UART, SPI, ADC UART, USART, I2C None UART UART, SPI UART, USB, I2C Data bus width 8-bits 8-bit 6800 or 6809 P 32-bit MFL5xxx P 32-bit MPC55xx P 32-bit ARM7 P 8-bit megaAVR P 8-bit PIC16 family Particulars

2 16-bit counters/time rs 512 bytes of EEPROM 2 16-bit timers MMU and DSP functionality 8-bit ADC, timers, PWM and watchdog 4 KB EEPROM, 10-bit ADC, PWM 8-bit ADC, 8-bit timer, comparat or Table 2. Embedded processor comparison chart Rami Abielmona SNODE Physical Characteristics (5) Operating system support The following characterize our ideal operating system

Multitasking and interrupt support Vast language and microprocessor support Ease of tool compatibility (compiler, assembler, ) Wide array of services (queues, semaphores, timers, ) Small area footprint (both program and data) Scaleable design Availability of debugging tools Standards compatibility Extensive device driver support Industrial and academic support! It is imperative to have a low interrupt latency, to allow for reentrancy and to support pre-emptive scheduling, as all will help us meet our real-time constraints when dealing with SNODE computational requirements Rami Abielmona SNODE Physical Characteristics (6) Rami Abielmona SNODE Physical Characteristics (7) RTOS Mentor Graphics Nucleus Cygnus Solutions eCos Lynx RealTime LynxOS Microsoft Corp. Windows CE QNX

Software QNX UC Berkele y TinyOS Avocet Systems AvSYS Micrium C/OS-II Target CPUs 68K, ARM, MIPS, x86, ColdFire, SPARC, H8, SH, TI DSPs ARM, MIPS, MPC8xx, SPARC, Toshiba TX139 68K, MIPS< MPC8xx, x86, SPARC, PARISC ARM , MIPS, PowerPC, SH, x86, Strong Arm, NEC MIPS, MPC8xx, x86 Network processor s

65816, 68HC08/1 1/12/16, 8051, Z8, Z80, 6809/01/0 3 ARM, AVR, Nios, x86, PowerPC, StrongARM , PIC-18xx, MIPS, 68K, MicroBlaze, Z80 Languages supported C, C++, Java Assembly, C, C++ Ada, assembly, C, C++, Java, Fortan, Perl Assembly, C, C++, Java Assembly, C, C++, Java nesC C C ROM footprint

< 1 Varies < 1 Varies 33 256 270 626 40 Varies 3500 0.8 640 2048 RAM footprint < 1 Varies < 1 Varies 11 115 40 720 Varies 4500 0.8 640 200 Multitaskin g Round robin, time slice, dynamic priorities Round robin, time

slice, fixed priorities Round robin, time slice, fixed priorities Round robin, time slice, dynamic priorities Round robin, time slice, dynamic priorities Priority schedulin g Time slice, fixed priorities Round robin, time slice, fixed priorities Licensing Per license Free Per license Per license Per license Free

Per license Free for research Particulars POSIX, TCP/IP, source code POSIX, TCP/IP, source code POSIX, TCP/IP, source code POSIX, TCP/IP, source code POSIX, TCP/IP, source code Eventbased POSIX, TCP/IP, source code POSIX, TCP/IP, source code Table 3. RTOS comparison chart Adapted from [15] Wireless technologies The following characterize our ideal wireless technology

Imperative low power utilization Simple transceiver circuitries Resilient to multipath effects Worldwide medium availability Standards compatibility Freely licensable Medium to wide range of operation Decent data transmission rate Industrial and academic support! Due to the limited power supply, researchers are trying to combine all three components (sensing, processing and transceiver) into tiny, low-power, low-cost units Rami Abielmona SNODE Physical Characteristics (8) Rami Abielmona SNODE Physical Characteristics (9) Wireless Technolog y BlueToot h IrDA IEEE 802.15.3a Ultrawideban d (UWB) IEEE 802.11a

IEEE 802.11b (Wi-Fi) IEEE 802.11g IEEE 802.15.4 (ZigBee) Data rate (Mb/s) 1-2 4 100-500 54 11 54 250 kbps and 20 kbps Output power (mW) 100 100 mW/ sr 1 40-800 200 65 30

Range (meters) 100 1-2 10 20 100 50 30 Frequency band 2.4 GHz Infrared 3.1-10.6 GHz 5 GHz 2.4 GHz 2.4 GHz 2.4 GHz and 868/915 MHz Comments 7 active nodes Very shortrange Low power,

short-range applications Wireless LANs with high data rate Wireless LANs with low data rate Wireless LANs with lower power Low dutycycle applications Table 4. Wireless technologies comparison chart Adapted from [16] SensIT program at DARPA [12] There are two main objectives to the program To develop novel networking techniques for SNETs deployed in unstructured and sometimes hostile environments To develop network information processing procedures, so as to extract useful, reliable and timely information from the SNET Data-centric routing focuses on the data generated by the sensors themselves, and avoids the overhead of assigning unique addresses to each SNODE The SNODEs are supposed to reach a networked consensus, when it comes to the application at hand (e.g. classification of a target) SensIT networks are interactive and can be dynamically tasked and queried by human operators, using a query/tasking language Multiple simultaneous users are allowed in the system Main four functions of SensIT are: detection, identification, location and tracking of objects Rami Abielmona

SNET Practical Implementation #1 Deployed the following equipment in the field 29 Palms 10 different armoured vehicles classified under 3 types Data analysis machines Various acoustic and seismic sensors organized in SNETs Goal was to classify each vehicle passing between two checkpoints using the gathered SNODE data Results were extremely good, except when a large number of vehicles or vehicles of various types were within the convoy Three different techniques were studied Winner Collaboration between 1 SNODE Collaboration between 2 SNODEs within target field of view Collaboration between 2 SNODEs not within target field of view Figure 5. SensIT scenario run-through Reproduced from [17] Rami Abielmona

SNET Practical Implementation #1 Smart Dust research project at UC Berkeley Also sometimes referred to as dust motes (small particle) There is one main objective of this project Miniaturization! The sub-units are to fit in a cubic millimeter There have been many variations of dust motes over the years Figure 7. RF mote Reproduced from [18] Figure 6. Laser mote Reproduced from [18] Figure 8. MEMS optical mote Reproduced from [18] Rami Abielmona SNET Practical Implementation #2 Figure 9. July 1999 mote Reproduced from [19] Figure 10. Ideal mote Reproduced from [19] Rami Abielmona SNET Practical Implementation #2 Intelligent Sensor Agent (ISA) research project at SITE (University of Ottawa) started in 2002 Environments range from the natural to the scientific to the military and even to the underwater There are many parameters to keep track of, and each parameter exhibits complex behavior. For example:

Chemical substances can be very tough to sense Sensors deployed in enemy territory could be destroyed The motivations are: Development of a new generation of autonomous wireless Robotic Intelligent Sensor Agents (R-ISAs) for complex environment monitoring Fusion of collected sensor data into a world model which is remotely available to human monitors Representation of the model in an interactive Virtualized Reality Environment The overall goal is to allow the human operator to remotely and continuously monitor the behavior of the environment and to actuate upon some of its constituents, if the need arises Rami Abielmona SNET Practical Implementation #3 There are many environmental properties [20] Accessible vs. inaccessible If accessible: can obtain complete and accurate information about environment Deterministic vs. non-deterministic If deterministic: can guarantee that a single action has a known effect on the environment Episodic vs. non-episodic If episodic: can link performance of the robot to a discrete set of episodes occurring in the environment Static vs. dynamic If static: can assume that environment does not change, unless it is from an action of a robot/agent Discrete vs. continuous

If discrete: can represent the environment with a fixed and finite number of actions and perceptions Most complex type of environment is the real world because it is inaccessible, non-deterministic, non-episodic, dynamic AND continuous! Rami Abielmona SNET Practical Implementation #3 Stationary Intelligent Sensor Agent Mobile Robotic Intelligent Sensor Agent Mobile Robotic Intelligent Sensor Agent Stationary Intelligent Sensor Agent Mobile Robotic Intelligent Sensor Agent Stationary Intelligent Sensor Agent WIRELESS COMMUNICATION NETWORK Head Mounted Display

DISTRIBUTED VIRTUALIZED REALITY ENVIRONMENT Haptic Feedback DISS is a distributed intelligent sensor system, utilized for environment monitoring. The concept involves both stationary ISAs and mobile ISAs and a wireless communication network. Distributed wireless network of mobile and stationary intelligent sensor agents deployed in the environment. The human operator monitors the environment from a remote location using interactive virtualized reality Figure 11. DISS architecture Reproduced from [21] Rami Abielmona SNET Practical Implementation #3 ISA Environment Composed of multiple robotic agents with onboard Sensing capabilities Computational entities Communication components

Environment has been mapped out (structured) and can be traversed following a localized map Wireless Communication Network Involves radio frequency technologies (commercial and academic) 802.11, Bluetooth, UWB Computational, incremental or approximate real-world model Conventionally accomplished with a computational model (i.e. Kalman filtering), but has since progressed to include biological models (i.e. neural networks) and probabilistic models (e.g. Bayesian estimates) Telepresence Remote access to the real-world model involves perception of visual, oral, haptic and maybe in the near future olfactory sensory information Distributed Multimedia Virtual Reality Environment Human Monitor (HCI) Figure 12. DISS flow diagram Rami Abielmona SNET Practical Implementation #3 DISS is composed of mISAs and sISAs, that each contain a small SNET on board. Its characteristics are as follows

Each agent has a global identification, but each sensor does not Utilizes peer-to-peer communication between the agents and broadcast within the SNET The number of agents is low, however the number of sensors is high Agents and SNODEs are limited in power, memory and computation Both network topologies are dynamic Low-level RF modulation transceivers and proprietary wireless protocols are utilized for the agents and their on-board SNETs UDP/IP is being used as the communication protocol between the agents We have chosen UDP/IP as the communication protocol amongst the agents, as that allows them to be queried by human users, and respond with assertions to those queries Micriums C/OS-II has been ported to the 68HCS12 microcontroller being used as the processing power of each agent A UDP/IP stack called lwIP [22] has also been ported to the same microcontroller Access points are being developed between the proprietary wireless protocols and standardized ones (e.g. ZigBee, Bluetooth and 802.11) Rami Abielmona SNET Practical Implementation #3 What should each mISA look like ? Each mISA contains a SNET composed of different types of sensors Ultrasonic ranging Exterior temperature Light intensity Smoke Pressure Each sensor leases computational

power, as well as communications packets, from the agents microcontroller and transceiver Each agent becomes the sink node for its on-board SNET The sink nodes can communicate with each other or with access points (APs) SNODEs are composed of Local sensory capabilities Distributed computational entities Shared communications timeslots Figure 13. Ideal mISA architecture Rami Abielmona SNET Practical Implementation #3 What does each mISA look like ? Figures 14-17. Current mISA architectures Rami Abielmona SNET Practical Implementation #3 Figures 18. ISA network architecture Rami Abielmona SNET Practical Implementation #3 SNETs are here to stay! Intelligent sensor systems will be prevalent in many of the everyday tasks that are either

Difficult for a human to achieve (e.g. fixing a part inside a car), or Mundane for a human to partake in (e.g. monitoring light intensity and closing the blinds accordingly) Future research directions can already be seen within the community Design of tiny, low-power, low-cost modules Network layer discovery and self-organization algorithms Collaborative signal processing and information synthesis Tasking and querying interfaces with the SNET Security for protection against intrusion and spoofing Reconfiguration techniques into suitable SNET configuration Rami Abielmona Research Directions We listed a few applications at the beginning such as Military (e.g. object tracking) Health (e.g. vital sign monitoring) Environment (e.g. natural habitat analysis) Home (e.g. motion detection) Are the following far off ? Wireless mobile SNETs can inform you of the availability of

a free parking spot (maybe even allow you to reserve it ?) Biological SNETs can monitor your health from within your body and can fight off viruses that may enter it Nanorobotic airborne SNETs can swarm towards disaster sites and traffic jams to give their respective audiences as much visual, and overall sensory, information as possible Rami Abielmona Potential Killer Applications Sensors are getting smaller in size and variable in nature Computing power is getting bigger and is being embedded Communications bandwidth is getting higher and transceivers are getting smaller Look for the following soon Negligible weight, dust particle size Integrated sensing/processing/communication Solar-powered modules Completely peer-to-peer topologies SNODEs will be like Oxygen [23] (MIT ubicomp project) SNETs will be embedded into the very fabric of our lives that they will inherently disappear! What about mobilizers and actuators ? Can you imagine smart dust particles that can mobilize and actuate upon their environment ? In approximately 5 years, on a PCB somewhere, after sensing and synthesizing (network concensus) that a micrometer-wide pin has been broken, a SNET self-organizes logically and physically and proceeds to solder the pin back to the chip! Rami Abielmona Market Trends

SNETs provide flexibility, fault-tolerance, high sensing fidelity, low-cost and rapid deployment Soon, we will be interacting with smart garments, smart appliances, smart sensor networks, and even smart floor tiles! However, we must keep in mind that we have to Fall back onto standards, when available Share information about our work Think passionately but design cautiously! Lots of work ahead of us! Lost of fun ahead of us! Work + Fun = Research Rami Abielmona Conclusions [1] C.E. Nishimura and D.M. Conlon, IUSS dual use: Monitoring whales and earthquakes using SOSUS, Mar. Technol. Soc. J., vol. 27, no. 4, pp. 13-21, 1994. [2] Proceedings of the Distributed Sensor Nets Workshop. Pittsburgh, PA: Dept. Computer Science, Carnegie Mellon University, 1978. [3] (1995) The cooperative engagement capability. [Online] Available: [4] (2003) CEC Provides Theater Air Dominance. [Online] Available: [5] I.F. Akyildiz, W.Su, Y. Sankarasubramaniam, E. Cayirci, Wireless sensor networks: a survey, Computer Networks, vol. 38, pp. 393-422, 2002. [6] S. Kumar and D. Shepherd, SensIT: Sensor information technology for the warfighter, in Proc. 4th Int. Conf. on Information Fusion, 2001, pp. TuC-1-3TuC1-9. [7] J.M. Kahn, R.H. Katz, K.S.J. Pister, Next century challenges: mobile networking for smart dust, Proceedings of the ACM, MobiCom99, Washington, USA, 1999, pp. 271-278. [8] (2003) AMPS -Adaptive Multi-domain Power aware Sensors. [Online] Available: [9] (2003) Intel Mote Exploratory Research in Deep Networking. [Online] Available: [10] Lars Erik Holmquist et al., Smart-Its Friends: A Technique for Users to Easily Establish Connections between Smart Artefacts, in Proc. Ubicomp 2001, Springer-Verlag LNCS 2201, pp. 273-291, 2001.

Rami Abielmona References (1) [11] (2003) Habitat Monitoring on Great Duck Island. [Online] Available: http:// [12] Chee-Yee Chong and Srikanta P. Kumar, Sensor Networks: Evolution, Opportunities and Challenges, in Proceedings of the IEEE, vol. 91, no. 8, August 2003. [13] Malik Tubaishat and Sanjay Madria, Sensor Networks: An Overview, IEEE Potentials, April/May 2003, pp. 20-23. [14] H.R. Everett, Sensors for Mobile Robots, A.K. Peters Ltd., MA: 1995. [15] (2003) Selecting a Real-Time Operating System. [Online] Available: [16] Steve Stroh, Ultra-Wideband: Multimedia Unplugged, IEEE Spectrum, Septebmer 2003, pp. 23-27. [17] (2003) Collaborative Classification Applications in Sensor Networks. [Online] Available: CollaborativeClassificationApplicationsInSensorNetworks_chartchai_SAM2002_08_07_02.p df [18] V. Hsu, J. M. Kahn, and K. S. J. Pister, "Wireless Communications for Smart Dust", Electronics Research Laboratory Technical Memorandum Number M98/2, February, 1998. [19] (2003) SMART DUST. [Online] Available: [20] Gerhard Weiss (editor), Multiagent Systems, The MIT Press, 2001. [21] E.M. Petriu, T.E. Whalen, R. Abielmona, "Autonomous Robotic Sensor Agents," Proc. ROSE03, 1st IEEE Intl. Workshop on Robotic Sensing 2003, Orebro, Sweden, June 2003. [22] (2003) lwIP: A Lightweight TCP/IP Stack. [Online] Available: http:// [23] (2003) MIT Project Oxygen: Pervasive Human-Centered Computing. [Online] Available: Rami Abielmona References (2) [1] Feng Shao and Leonidas Guibas, Information Processing in Sensor Networks: Second International Workshop, Ipsn 2003, Palo Alto, Ca, Usa, April 22-23, 2003: Proceedings (Lecture Notes in Computer Science, 2634), Springer-Verlag Publication, July 2003. [2] Edgar H., Jr. Callaway and Edgar H. Callaway, Wireless Sensor Networks: Architectures and Protocols, CRC Press, August 2003. [3] Victor Lesser, Charles Ortiz and Milind Tambe, Distributed Sensor Networks: A

Multiagent Perspective, Kluwer Academic Publishers, October 2003. [4] Anna Hac, Wireless Sensor Network Designs, John Wiley & Sons, January 2004 Rami Abielmona List of Recent Books Questions ? Comments ? Rami Abielmona SITE Ph.D. Candidate SMRLab Research Assistant [email protected] Rami Abielmona Contact Information

Recently Viewed Presentations

  • Camera API - Horstmann

    Camera API - Horstmann

    Camera API - Capture PhotoCreate a Preview Class and Preview Layout. A Preview Class is a SurfaceView that can display live image data coming directly from the Camera. Create a class that extends SurfaceView and implements Surfaceholder.Callback..

    Gifted Behaviors Commentary (GBC) Is highly reflective and/or sensitive to his/her environment. Readily learns and adapts to new cultures . Is acquiring language at a rapid pace . Exceptional ability to learn. Uses and interprets advanced symbol systems in academics,...
  • CS 115 Lecture 6

    CS 115 Lecture 6

    CS 115 Lecture . Graphics Part I. Taken from notes by Dr. Neil Moore. The graphics library. So far all our programs have interacted with the user through standard input and output (keyboard and shell window) ... Keep it open...
  • Cost Transfers: How to Make Them Work for You and Not Against You

    Cost Transfers: How to Make Them Work for You and Not Against You

    Basic "Textbook" Definition. A . Cost Transfer. is the transfer of an expenditure initially posted. to one project and then transferred to another project. ... Transfers within the same project / award may have budget period restrictions. all transactions under...
  • ESA Collaborative Environment for Cal/Val Andrea Della Vecchia,

    ESA Collaborative Environment for Cal/Val Andrea Della Vecchia,

    Joint ESA-NASA MAP Objectives. To provide to the users a common virtual working environment. Federation of two (or more) platforms . to allow to both US and European users a transparent access to the content of both platforms,
  • Arrow&#x27;s Theorem

    Arrow's Theorem

    Arrow's Theorem. Transitive and unrestricted. Transitive and complete. Independence of Irrelevant Alternatives and Unanimity condition. Alternative reading: All democratic voting system will fail to satisfy at least one of independence of irrelevant alternatives, the unanimity condition or transitivity of the...
  • Public Election Fund Compliance Seminar Resources: Summary Guide

    Public Election Fund Compliance Seminar Resources: Summary Guide

    The allowable uses of money in a publicly funded campaign account are limited to expenses directly related to election campaign activities and expenses incurred during the specific time period for either the primary or general election.
  • SETPOINT Berkshire - HISTORY

    SETPOINT Berkshire - HISTORY

    How to be synoptic. Those who write synoptically include references to various sections (or rulers) of the 100 year period in each paragraph. Better candidates make cross references, comparisons and links between different periods in each paragraph