|
|
|
|
|
Graduate Research Assistant,
Applied Intelligent Systems Laboratory (AISL) |
- Development of a
Mathematical Model for Energy Management in Wireless Sensor Networks (WSNs)
|
|
Wireless sensor networks (WSNs)
are collections of a large number of sensor nodes. Each
node is |
|
powered by a finite energy source, like a battery pack.
To reduce the maintenance costs, it is important |
|
to manage and minimize the energy consumption of the
network by optimizing battery lifetime and to |
|
better
estimate proper battery replacement time for each node. |
|
- Intelligent and Multi-resolution Signal
Decomposition for Classification of Power Quality Disturbance
|
|
A MATLAB toolkit is developed to decompose nonlinear and non-stationary power
quality disturbance
|
|
signals.
The multi-resolution signal processing tools; wavelets
and empirical mode decomposition
|
techniques. The decomposed signals are used to determine
the feature space and are classified using |
|
a nonlinear multi-class support vector machines. |
|
|
|
|
- Non-Parametric
Regression Methods for Quantification of Reactor Safety Margin
|
|
In
collaboration with Prof. M. L. Bertadano (School of
Nuclear Engineering), a non-parametric research
|
|
|
regression
model was developed to perform the sensitivity analysis
of input parameters and a realistic |
|
values for safety margins were determined by evaluating
the uncertainty of critical input parameters of |
|
|
- Kernel
Regression Approach to Short -Term Load Forecasting
|
|
Investigated and evaluated the application of kernel regression (nonparametric)
approach to short term |
|
load forecasting. Gaussian kernel was used
and the bandwidth of the kernel was selected using
Direct |
- Short Term
Load Forecasting using Artificial Neural Networks
|
|
The performance of the recurrent neural network (RNN)
and multilayer perceptron (MLP) neural network |
|
for short term load forecasting using previous load history and
temperature information was compared. |
|
Research Associate,
Hardcopy Technologies Laboratory |
|
Developed a multi-projector image blending scheme
independent of the projector orientations and also |
|
developed a Matlab toolbox for PR - 705 spectroscan. |
|
Graduate Research Assistant,
Imaging, Robotics, and Intelligent System (IRIS)
Laboratory |
-
Illumination Chromaticity Estimation via Kernel Regression
|
|
Proposed a simple nonparametric approach known as kernel
regression to estimate the illumination ch- |
|
romaticity. The performance of kernel
regression was compared with neural networks and support
vec- |
- Ridge
Regression Approach to Color Constancy
|
|
Estimated the illumination chromaticity using a linear
machine learning technique, known as, ridge regr- |
|
ession and showed that it performed better
than neural networks and support vector machines (SVM). |
|
The uncertainty analysis using
bootstrapping showed that ridge regression and SVM are
more stable |
|
than neural networks
in estimating chromaticity. |
- Image
Restoration using L1 Norm Penalty Functions
|
|
Least Absolute Shrinkage and Selector Operator (LASSO)
an efficient statistical modeling technique |
|
was extended to image restoration. The
performance of LASSO is compared with
established Total
|
|
Variation (TV) image restoration
technique. LASSO outperformed TV restoration in terms of
computatio- |
|
nal time and
achieved better restoration than TV. |
- Discrete
Nonlinear Programming for the Optimal Selection of LED Arrays
|
|
In collaboration with Siemens Energy and Automation,
Johnson City, TN, formulated an algorithmic |
|
approach for the optimal selection of set
of Light emitting diodes (LED) from a 260
combination of LED |
|
bins in each
color. An automation of the placement of the selected
LEDs in the circuit as per design cr- |
|
|
