Optimizing Sensor Network Boundary Estimation and Localization Algorithms for TMS320C6xOutlineMotivationBoundary EstimationK-mean clusteringSupport Vector MachinesSimulationsLocalizationProject ScopeApproachSample CCS SnapshotResult previewConclusionOptimizing Sensor NetworkBoundary Estimation and Localization Algorithms for TMS320C6xKamal SrinivasanNiveditha SundaramOutline•Motivation•Background–Boundary Estimation–Localization•Project Scope•Approach•ConclusionMotivation•Sensor Networks: Network of wireless enabled sensors•Applications–Environmental monitoring•Constraints–Delay Sensitive–Small memory footprintBoundary Estimation•Estimating boundary of a contaminated field•Oil spill•Biochemical attack•Radioactive and hazardous gas leaks•Algorithms•K-mean•Support Vector Machine (SVM)K-mean clusteringIterati vely,-Partition data into K=2 subsets based on initial seeds- Calculate centroid locationsBoundary Point DetectionthresholdfavgfavgiCiiCi)()(21Support Vector Machines-Partition data using 2 support lines that maximizes the distance between the 2 data setsSupport line equations-w1x + w2y + b = ±1 Boundary line detection-w1x + w2y + b = 0SimulationsEstimation using SVM for a rectangular boundaryLocalizationL(i) of node i(lxi, lyi)(uxi, uyi)(lxA, lyA)(uxA, uyA)NewL(A)F(i) of node iL(i) : location rectangle of node iF(i) : neighbor rectangle of node iLocation Estimation using neighborsProject Scope•Optimizing for TMS320C6x DSP processor•TMS320C6x architecture–8-way VLIW architecture–8 functional unitsApproach•Understanding the dependencies of the algorithm – DG•Software ILP•Loop Transformations •Loop Unrolling•Loop Interchange•Avoiding function call within loops – use function pointers•Tool – Code Composer StudioSample CCS SnapshotResult previewFunction() Speedupfindrect() 0.8906checkrect() 0.8741updateL() 0.9158updateG() 0.8266updateF() 0.9172Conclusion•Analyzed sensor network boundary estimation and localization algorithms•Code optimization is useful for sensor
View Full Document