AE Seminar: Model-Based Estimation and Characterization of Collective Motion in Animal Groups

Dr. Sachit Butail from Dynamical Systems Laboratory, Polytechnic Institute of New York University, would be presenting a seminar through Skype, as per the following schedule.
------------------------------------------------------------------------
Title: Model-Based Estimation and Characterization of Collective Motion in Animal Groups
Speaker:   Dr. Sachit Butail
               Dynamical Systems Laboratory,
               Polytechnic Institute of New York University

Date  :     Friday, November 9, 2012
Time  :     05:00 P.M.
Venue :    AE conference room
------------------------------------------------------------------------
Abstract:
------------
In this talk I will first describe application of nonlinear estimation and computer vision methods for the design of multi-target tracking systems to study collective behavior in animal groups. The tracking systems are currently being used to study information transmission in schooling fish, and swarming and mating behavior of wild malarial mosquitoes. For tracking fish, I will present methods to automatically initialize, predict, and reconstruct shape trajectories of multiple fish through occlusions. For tracking mosquitoes, which appear as faded streaks on in-field footage, I will present novel likelihood functions and algorithms to extract velocity information from the streaks, adaptively seek missing measurements, and resolve occlusions within a multi-hypothesis framework. In each case the research has yielded an unprecedented volume of trajectory data for subsequent analysis.
Moving forward, I will discuss utilization of the three-dimensional mosquito trajectory data to mathematically characterize the swarming events while indicating the effects of ambient high wind. For an individual mosquito, empirical fits to a damped harmonic oscillator model driven by white noise disturbance are used to differentiate horizontal and vertical motion. For the swarm, average disagreement in direction of motion between individual mosquitoes is used to show decreasing level of coordination with successive nearest neighbors. The broader impact of this work is to advance the understanding of animal groups for the design of bio-inspired robotic systems, where, similar to the animal groups we study, the collective is able to perform tasks far beyond the capabilities of a single inexpensive robot.

Brief Bio:
----------
Dr. Sachit Butail is a postdoctoral fellow in the Dynamical Systems Laboratory at the Polytechnic Institute of New York University, where he is working on machine learning algorithms to study animal-robot interactions. He completed his Ph.D. in June 2012 in Aerospace Engineering from the University of Maryland, College park, where his dissertation was on motion reconstruction of animal groups using methods from estimation theory and computer vision. He received a Masters in Systems Engineering with a project on Spacecraft design from Cornell University and a Bachelors in Mechanical Engineering from Delhi College of Engineering, Delhi University. His research interests are in dynamical systems modeling, collective behavior, and machine learning. He is a member of IEEE and SIAM.