Efficient Sensing using Mobile Robots – From Theoretical Foundations to in-field Validation

Amarjeet Singh
Advisor: William J. Kaiser

Wednesday, March 11, 2009 at 10:00am-11:00am
Engr IV Room 57-124

Abstract:
In the development of sensor network technology, one of the most long-standing research challenges has been the development of methods that efficiently monitor large (possibly uncertain) environments with complex dynamics that can only be partially observed. Often with limited sensing resources, mobile robots are used to achieve sufficient spatial and temporal coverage. However, planning the motion of these robots - coordinating their paths in order to maximize the amount of information collected while placing bounds on their resources (e.g., path length, energy capacity or sensing time) - is a NP-hard problem.

In this talk I will present several novel algorithms for performing robotic path planning in such complex environments, in order to maximize the “informativeness” of the locations visited by the robots. We provide strong theoretical guarantees for each of the proposed algorithms by exploiting different intuitive properties - diminishing returns, locality and monotonicity, of the objective function to optimize while maximizing the collected information. We also analyze the benefit of trade-off between exploration (gathering information about the environment) and exploitation (using the current knowledge about the environment most effectively) using an objective function that simultaneously optimizes exploration and exploitation. Using this objective function, we perform path planning in uncertain environments non-myopically i.e. planning ahead for observations that can be made in the future. We show validation of each of the proposed algorithms on real world sensing datasets as well as from several field experiments in critical environmental monitoring applications.

Biography:
Amarjeet Singh completed his undergraduate education from Indian Institute of Technology, Delhi in 2002. From 2002 – 2004, he worked as Senior Research and Development Engineer at Tejas Networks, Bangalore, India. In 2005, he joined Center for Embedded Networked Sensing (CENS), UCLA and completed his MS in Electrical Engineering in 2007. He was awarded the Edward K. Rice outstanding MS student award in 2007. He is also a recipient of Dissertation year fellowship for 2008-09. Part of the thesis work was completed during research collaborations at Carnegie Mellon University and Sydney University.