Shivakar Vulli

Research

Areas of Interest

I worked as an ASTRO (Advanced Short-Term Research Opportunity) appointee at Oak Ridge National Laboratory with Computational Earth Sciences group in the Computer Science and Mathematics Division. My research at ORNL involved applying and improving existing multi-variate cluster analysis methodologies, combined with other statistical and data mining techniques to facilitate model-data comparison for observations from DOE’s atmospheric radiation measurement program.

As a graduate research assistant at the Missouri University of Science and Technology, I am worked with the Airborne Reconnaissance and Image Analysis (ARIA) Lab and Intelligent Systems Center (ISC). During the course of my study at Missouri S&T, I worked with several multidisciplinary teams in the areas of robotics, airborne mine and minefield detection, evolutionary computation and artificial life.

Broadly, my areas of interest include High Performance Computing, Parallel and Distributed Computing, Multicore Programming, Evolutionary Computation, Robotics, Machine Learning and Data Mining. Particularly, I am interested in modeling and simulation of complex systems, novel applications of evolutionary and other bio-inspired algorithms to data mining & robotics, and applications of High Performance Computing.

Before joining ARIA, I worked on Computer Aided Design and Engineering (CAD/CAE) at V.R. Siddhartha Engineering College, India, which mainly involved Computer Aided Finite Element Analysis of Laminated Composites.


Research Projects


An Artificial Life Approach to Evolutionary Computation: From Mobile Cellular Algorithms to Artificial Ecosystems (Computer Science Master’s Thesis. Funded by ARIA Lab)

Developed robust agent-based evolutionary algorithms inspired by principles of artificial life and emergence to solve optimization problems with high degree of epistasis and multimodality.

  • Introduced a new class of spatially structured evolutionary algorithms called Mobile Cellular Evolutionary Algorithms (MCEAs), characterized by evolving individuals moving in a spatially structured population.
  • Investigated the effects of mobility and population density on the selection pressure and performance of mobile cellular evolutionary algorithms.
  • Introduced a new architecture, called Artificial Ecosystems (AES) for developing agent-based evolutionary algorithms.
  • Developed two population dynamics schemes for regulating the reproduction of agents in Artificial Ecosystems.
  • Demonstrated the efficacy of the MCEA and AES methodologies for multi-modal optimization using a test suite of problems.


Individual-Based Artificial Ecosystems for Design and Optimization (Mechanical Engineering Master’s Thesis. Funded by ARIA Lab and ISC)

An individual-based (agent-based) framework inspired by naturally occurring ecosystems to solve engineering design and optimization problems. A given design or optimization problem is “mapped” to the framework, which is them simply run. The solution to the given problem “emerges” through the interactions of the individuals (agents) among themselves and with their environment.

A model in this framework is created by first identifying a natural ecosystem (currently at the discretion of the modeler) capable of solving the given problem. The problem is then encoded into the selected ecosystem. Any modifications to the ecosystem which might improve the computational efficiency of the model are applied. The completed model is then run to obtain the solution to the problem.

  • Developed an individual-based (agent-based) framework inspired by naturally occurring ecosystems for solving engineering design and optimization problems.
  • Developed individual level behaviors comparable to biological processes.
  • Demonstrated the emergence of ecosystem level phenomena such as population dynamics, niche formation, adaptation to environment from individual-level interactions.
  • Demonstrated the efficacy of the approach by the use of a visually illustrative example of parameter estimation for binary texture synthesis.


Multi-Variate Cluster Analysis (Funded by Forrest Hoffman and William Hargrove at ORNL through the ASTRO Program)

Applying and improving multi-variate cluster analysis methodologies to facilitate model-data comparison for observations from DOE’s atmospheric radiation measurement (ARM) program.

  • Investigated the performance of several clustering algorithms including k-means, k-center, particle swarm optimization, and their variants for clustering large datasets (~7 million to ~1 billion observations).
  • Investigated the effect of parallel I/O settings on the performance of large number of random read and write access pattern which the implementation requires.
  • Investigated the effect of various seeding techniques on the clustering performance of k-means algorithm.
  • Investigated the scalability of the implementations up to 10,000 processor cores.
  • Developed tools (serial and parallel) to facilitate analysis of clustering results, including, data preprocessor, cluster quality analyzer, random sampler, custom timing library and random I/O analysis.


Airborne Mine Detection (Funded by ARIA Lab)

Developed algorithms and tools for image-to-image registration and image-to-ground geo-referencing and anomaly detection for mine and minefield detection from airborne infrared and multi-spectral imagery.

  • Conducted an extensive survey to assess current open source geospatial solutions for a real-time web-based geospatial analysis system.
  • Implemented feature extraction, edge detection and image segmentation algorithms to improve the performance of an existing Matlab application for semi-automatic image registration.
  • Implemented spatial and temporal transformation algorithms to improve the performance of an existing Matlab application for geo-referencing airborne imagery.
  • Developed algorithms for feature extraction and feature quality analysis for automatic image registration.
  • Developed a Matlab application for automatic image registration using feature extraction and image metadata.
  • Developed a distributed image registration application using Matlab Distributed Computing Toolbox for heterogeneous Linux clusters.
  • Developed a Matlab application for automatic geo-registration of airborne imagery using image metadata, ground control point information and bundle adjustment.
  • Developed a C++/IPP (Intel Performance Primitives) application for automatic image registration using phase correlation and optical flow.
  • Developed a Qt/OpenGL front-end GUI for the image registration tool.
  • Developed a C++ wrapper to interface existing Matlab applications with NITF (National Imagery Transmission Format) imagery.
  • Developed  an in-memory JPEG 2000 decoder using JasPer library to interface existing Matlab applications with JPEG 2000 compressed imagery.
  • Developed C++/IPP and IDL applications for correcting band-to-band misalignment in multi-spectral imagery.
  • Developed a C++/IPP application for anomaly detectino in single and multi-band imagery using RX (Reed and Xioli) detector.


False Alarm Mitigation in Airborne Minefield Detection (Funded by ARIA Lab)

Worked in the capacities of database designer, administrator and application developer, in a multi-institutional effort to develop a first of its kind, geo-spatial database of spectral signatures of false alarms in airborne mine and minefield detection systems.

  • Contributed to the design and implementation of a PostgreSQL/PostGIS database for false alarm library.
  • Developed and maintained a MEX connectivity program using the libpq library for querying PostgreSQL database from Matlab.
  • Developed a Matlab application for automatic extraction of false alarms.
  • Designed and developed a PostgreSQL database to store spectral and spatial information for the identified false alarms.
  • Developed and optimized queries for accessing image analysis information from the database.


Directional Vision Based Multi-Robot Formation Control (Funded by ISC)

A decentralized behavior based framework for multi-robot formation control using only directional vision. For more information on this project please refer project member Gerard Sequeira’s M.S. Thesis.

  • Developed  custom firmware for the Surveyor SRV-1 robot platform which enable d the robots to navigate autonomously using only infrared sensors, send captured images to a remote host for processing and receive navigation information form the remote host.
  • Developed a subsumption architecture based framework which allowed the robot platform to autonomously avoid obstacles using infrared sensors at the lowest level. Path planning and navigation was accomplished by processing vision data at a higher level.
  • Implemented serial communications routines for message passing between the remote host and the mobile robots via Zigbee radio.
  • Developed an in-memory JPEG decoder using JasPer library to improve the performance of the remote host.
  • Improved image analysis performance on the remote hsot using OpenMP parallelization.


Real-Time Traffic Simulation (Funded by ISC)

Developed a scaled simulation environment for studying the performance of autonomous navigation algorithms for DARPA’s Urban Challenge competition.

  • Developed a C++/OpenCV based front-end for real-time remote visualization.
  • Determined the location and orientation of each car in the simulation arena via color contour detection from images captured using an overhead camera.
  • Developed serial communications routines for sending necessary navigation information from the remote host to on-board microcontrollers via radio.


Papers and Presentations


Journal Publications

  • Vallabhaneni, B. K. M., Vulli, S. S., and Kondapalli, P. K., “Prediction of Engineering Properties of GFRP lamina: Tensile behavior,” Journal of Institution of Engineers (India), Vol 86, 2006, pp191-194.

Conference Publications

  • Hoffman, F. M., Mills, R. T., Kumar, J., Vulli, S. S., Hargrove, W. W., “Geospatio-temporal Data Mining in an Early Warning System for Forest Threats in the United States, ” in the Proceedings of the 2010 IEEE International Geoscience and Remote Sensing Symposium (IGARSS 2010, July 25-30, Honolulu, Hawaii), July 2010.
  • Singiresu, D., Agarwal, S., Vulli, S., Ramakrishna, H, N., “GPU Based Processing for Airborne Detection,” in the Proceedings of the SPIE 2010, Defense, Security and Sensing Conference, April 2010.
  • Agarwal, S., Vulli, S., Malloy, N., Lord, E., Fairley, J., Sabol, B., Johnson, W., Ess, R., Trang, A., “Collection and Evaluation of False Alarm Signatures in Background Data,” in the Proceedings of the SPIE 2009, Detection and Sensing of Mines, Explosive Objects and Obscured Targets XIV, April 2009.
  • Sanka, A., Vulli, S., Agarwal, S., Ess, R., Trang, A. H., “Spectral and Spatial Analysis of False Alarms in Background Data,” in the Proceedings  of SPIE 2009, Detection and Sensing of Mines, Explosive Objects, and Obscured Targets XIV, April 2009.
  • Vulli, S. S., and Agarwal, S., “Individual-Based Artificial Ecosystems for Design and Optimization,” in the Proceedings of the Genetic and Evolutionary Computation Conference 2008, GECCO ‘08, pp. 273-280, 2008.
  • Vallabhaneni, B. K. M., Vulli, S. S., Sudhakar, A.V., and Rao, U. K., “Finite Element Modeling and Simulation for the Shear and Thermal Behavior of FRP Lamina,” in the Proceedings of International Conference on Emerging Adoptive Systems and Technologies (EAST-2005), organized by Noorul Islam College of Engineering, Kumaracoil, Tamilnadu, Dec. 16-17, 2005, pp 966-975.
  • Vallabhaneni, B. K. M., Vulli, S. S., Rao, U. K., and Katta, M. R., “Prediction of Shear and Thermal properties of GFRP Laminates,” in the Proceedings of International Conference on Total Engineering, Analysis & Manufacturing Technologies (TEAM TECH2006), IISc Bangalore, Feb 28th - March 2nd 2006, pp 18-19.
  • Vallabhaneni, B. K. M., Vulli, S. S., Rao, U. K., and Katta, M. R., “Prediction of Mechanical Properties of FRP Laminates Using Finite Element Method,” International Conference on Global Manufacturing and Innovation (GMI), being organized by Coimbatore Institute of Technology jointly with University of Massachusetts, U.S.A., Coimbatore during 27-29 July 2006.
  • Pingali, V. V., Vulli, S. S., Sudhakar, A.V. and Vallabhaneni, B. K. M., “Prediction of Elastic Constants of Angle-Ply Lamina,” in the Proceedings of the National Conference on Recent Trends in Mechanical Engineering (RTIME – 2004) July 23-24, 2004, SRES College of Engineering, Kopargaon. pp. 32.

Symposiums

  • Vulli, S. S., and Agarwal, S., “Artificial Ecosystems: An Artificial Life Approach to Multiobjective Optimization,” SwarmFest 2009, June 2009.
  • Vulli, S. S., and Agarwal, S., “An Individual-Based Predator-Prey Ecosystem for Image Segmentation: Effect of Explorers,” SwarmFest 2008, May 2008.
  • Vulli, S. S., and Agarwal S., “Artificial Life as a Framework for Systems Design,” ISC Research Symposium, Missouri S&T, Rolla, Apr. 2008.
  • Vulli, S. S., Sequeira, G. D., Agarwal S. and Krishnamurthy K., “Directional Vision Based Multi-Robot Formation Control,” ISC Research Symposium, UMR, Rolla, Apr. 2007.
  • Vulli, S. S., “Mechanical Properties of a Unidirectional Fiber-reinforced Angle-ply Lamina,” TECHNOFEST-2003, VRSEC, Vijayawada, Dec. 2003.
  • Vulli, S. S., “Design for Assembly Using Feature Technology,” 6th National Convention of ISTE Students, JNTUCE, Hyderabad, Oct. 2003.

Presentations

  • Artificial Life as a Framework for System Design, ISC Poster Presentations, UMR, Rolla, Apr. 2007.
  • Visual Perception and Navigational Learning for LAGR, ISC Poster Presentations, UMR, Rolla, Oct. 2006.
  • A Real Time Traffic Simulation for DARPA Urban Challenge, ISC Poster Presentations, UMR, Rolla, Oct. 2006.