4003-543/4005-742 Ad Hoc Networks Module 7. Sensor Networks -- Lecture Notes Prof. Alan Kaminsky Rochester Institute of Technology -- Department of Computer Science Examples of Sensor Nodes Sensor nodes used in the Information Sciences Institute (ISI) Laboratory for Embedded Networked Sensor Experimentation (I-LENSE) project Sensor nodes: http://www.isi.edu/ilense/testbed/index.html I-LENSE web site: http://www.isi.edu/ilense/index.html ISI web site: http://www.isi.edu/ ISI is part of the University of Southern California: http://www.usc.edu/ These all use either wireless Ethernet (802.11a, 802.11b) or specialized radios for communication; frequencies 400 MHz, 900 MHz, 2.4 GHz, 5 GHz; outdoor range 500-1000 feet   Wildfire sensor nodes used by Bob Kremens, Senior Research Scientist, Center for Imaging Science, RIT Bob Kremens's web site: http://www.cis.rit.edu/~rlkpci/ Uses the 2-meter amateur radio band for communication, frequency 150 MHz, outdoor range many kilometers Uses either data communication (modem) or voice communication (speech synthesizer) over the analog radio channel Sensor Network Applications Sensing modalities   Applications Sensor Network Topologies Sensor nodes vs. query (human interface) nodes   Ad hoc mesh topology (short range radios)   Star topology with base station (long range radios)   Hierarchically clustered topology (hybrid) Many smaller, low-power nodes with short range radios A few larger, higher-power nodes with short and long range radios Querying and Data Fusion Queries tend to be high-level "Where is the wildfire?" "Where is the tank and what direction is it moving?"   Sensor data tends to be low-level "The temperature at location (x,y) is T degrees" "The acoustic signal strength at location (x,y) and time t is S dB"   How to bridge from sensor data to query results?   Alternative: Send all sensor readings to a base station (computer), base station computes results Pros: Sensor nodes can be "dumb," only base station needs to be "smart" Cons: Uses much battery power transmitting and forwarding messages over many hops, therefore short battery life and short network life The node's wireless interface can consume 1,000 to 10,000 times as much power as the node's CPU   Alternative: Data fusion -- Sensor nodes compute query results amongst themselves, only one node (or a few nodes) sends the results to a base station Pros: Much less power consumed sending messages, therefore much longer battery life and network life Cons: Nodes must be smarter and do more processing Still, it's a good tradeoff Data Fusion Example: Target Counting F. Zhao, J. Liu, J. Liu, L. Guibas, and J. Reich. Collaborative signal and information processing: an information-directed approach. In S. Iyengar and R. Brooks, editors, Distributed Sensor Networks (Chapman & Hall/CRC, 2005), pages 185-200.   Query: How many targets are within the sensor network field?   Sensing modality: Acoustic signal strength   Algorithm Each node measures the acoustic signal strength S Each node broadcasts its measurement to the neighboring nodes Each node waits until all its neighbors have reported (or until a certain time has elapsed) If a node has the largest value of S among all its neighbors, the node decides it is the closest-to-target node Each closest-to-target node sends a target report to the base station The number of target reports is the answer to the query   This resembles a distributed leader election algorithm The node with the highest S value becomes the leader among its neighbors   Variations Number and locations of targets within the sensor network field Signal strength contours Data Fusion Example: Target Tracking D. Friedlander. Parameter estimation. In S. Iyengar and R. Brooks, editors, Distributed Sensor Networks (Chapman & Hall/CRC, 2005), pages 573-596.   Query: What is the target's location and velocity?   Sensing modality: Acoustic signal strength   Algorithm Each node measures the acoustic signal strength S as a function of time t Each node determines the target's closest point of approach (CPA) CPA = the time at which S(t) is a maximum Each node broadcasts to the neighboring nodes: S and t measurements at the CPA Node's (x,y) location Each node waits until all its neighbors have reported (or until a certain time has elapsed) If a node has the largest value of S among all its neighbors, the node decides it is the closest-to-target node (same as previous example) The closest-to-target node estimates the target's motion in the X direction Take all the (t,x) readings from the neighboring nodes Do a linear regression to estimate x(t) = a t + b The closest-to-target node estimates the target's motion in the Y direction Take all the (t,y) readings from the neighboring nodes Do a linear regression to estimate y(t) = c t + d The closest-to-target node sends the estimated parameters (a,b,c,d) to the base station The base station now knows x(t) and y(t) The base station can track the target's location as a function of time The base station can determine the target's velocity (speed and direction) Ad Hoc Networks • 4003-543-01/4005-742-01 • Spring Quarter 2007 Course Page Alan Kaminsky • Department of Computer Science • Rochester Institute of Technology • 4486 + 2220 = 6706 Home Page Copyright © 2006 Alan Kaminsky. All rights reserved. Last updated 15-May-2006. Please send comments to ark­@­cs.rit.edu.