New version page

NSFRPI_selfOpt

Upgrade to remove ads

This preview shows page 1-2-3-4-5 out of 14 pages.

Save
View Full Document
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 14 pages.
Access to all documents
Download any document
Ad free experience

Upgrade to remove ads
Unformatted text preview:

NSF-RPI Workshop on Pervasive Computing and Networking, April 29-30, 2004 1Self Optimization inWireless Sensor NetworksBhaskar KrishnamachariAutonomous Networks Research GroupDepartment of Electrical Engineering-SystemsUSC Viterbi School of Engineeringhttp://ceng.usc.edu/~anrg2Sensor networks represent a fundamental paradigm shift from InterpersonalCommunication to Communication with the EnvironmentThis must change the way we analyze and design these systems.3Analysis and Designof Wireless Sensor Networks• Mathematical Models and Performance Analysis• Self-optimizationProtocol• algorithm choice• parameters Application• traffic • placement• topology Environment• channel condition• spatial correlation• data dynamics4Analysis and Designof Wireless Sensor Networks• Mathematical Models and Performance Analysis• Self-optimizationProtocol• protocol selection • parameter selectionApplication• traffic • placement• topology Environment• channel condition• spatial correlation• data dynamics5Analysis of Directed Diffusion• Expressions for overhead of push versus pull diffusion to determineapplication conditions where each is suitableKrishnamachari, Heidemann, “Application-Specific Modeling of Information Routing in Wireless Sensor Networks,” IEEE IPCCC-MWN ‘04.6Analysis of ACQUIRE• Analysis of an active query forwarding mechanism providing a lookaheadparameter to tune between flood-based and trajectory-based queryingSadagopan, Krishnamachari, Helmy, “Active Query Forwarding in sensor networks,” Ad Hoc Networks,2004 (to appear).trajectoryflood7Analysis of Routing with Compression• There exists a static clustering technique that provides near optimal routingwith compression across a wide range of spatial correlation levelsPattem, Krishnamachari, Govindan, “Impact of Spatial Correlation on Routing with Compression inWireless Sensor Networks,” ACM/IEEE IPSN, 2004 [best paper award].8Analysis and Designof Wireless Sensor Networks• Mathematical Models and Performance Analysis• Self-optimizationProtocol• protocol selection • parameter selectionApplication• traffic • placement• topology Environment• channel condition• spatial correlation• data dynamics9Self-Optimization*• The traditional approach is to take into account application requirementsprior to operation and pre-configure/pre-optimize the protocol parameters.• This is simply insufficient in sensor networks, where the characteristics ofthe environment, which can be inherently unpredictable, play a key role.• A powerful design principle we must embrace is the use of autonomouslearning through sensor observations during operation, so that networkprotocols can optimize their own performance over time.• These protocols must be distributed and localized for scalability andefficiency.* Ongoing collaboration with Marc Pearlman, Kraken Inc.10Simultaneous Localization and Tracking• Current techniques attempt to localize nodes prior to operation based oncommunication constraints or ranging estimates• In a target tracking application, sensor observations of the moving targetintroduce additional constraints that can be used to further reduce thelocalization error over timeAram Galstyan, Bhaskar Krishnamachari, Kristina Lerman, and Sundeep Pattem, "Distributed OnlineLocalization in Sensor Networks Using a Moving Target," ACM/IEEE IPSN, 2004.11Model-based Compression• In some cases, the underlying physics of the phenomena can provide aspatio-temporal model (e.g. PDE models for heat or chemical diffusion).• Then instead of sending raw data, it suffices to send the model parameters,which can be learned through observations over timeLorenzo Rossi, Bhaskar Krishnamachari, C.-C. Jay Kuo, "Modeling of Diffusion Processes with PDEModels in Wireless Sensor Networks," SPIE Defense & Security Symposium, 2004.12Reinforced Querying and Routing• Technique suitable for querying about targets or pushing data to sinks thathave a underlying (unknown) probabilistic location pattern• The basic idea is to start with a random walk, and change forwardingweights based on reinforcementsKrishnamachari, Zhou, Shademan, “LEQS: Learning-based Efficient Querying for Sensor Networks”,USC CS Technical Report, 03-795, 2003.13Reinforced Querying and Routing• Converges over a period of time to an efficient trajectoryKrishnamachari, Zhou, Shademan, “LEQS: Learning-based Efficient Querying for Sensor Networks”,USC CS Technical Report, 03-795, 2003.14Summary and Conclusions• Sensor Networks represent a fundamental paradigm shift from inter-personal communication to communication with the environment. This hassignificant implications for both analysis and design:• Analysis: Protocol performance must be analyzed with respect to acombination of environmental effects, application specifications and protocolparameters.• Design: Protocols must be designed to be self-optimizing, improvingautonomously over time by incorporating sensor


Download NSFRPI_selfOpt
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view NSFRPI_selfOpt and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view NSFRPI_selfOpt 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?