[Brown CS Talks] Michael Benjamin Dissertation Defense in Lubrano on 3/15/02 at 10 am

[talks-admin@list.cs.brown.edu]

		  The Department of Computer Science
			   BROWN UNIVERSITY

			      
			       presents

 
			   Michael Benjamin

		      Ph.D. Dissertation Defense


		 Friday, March 15, 2002  at 10:00 am
	       Lubrano Conference Room (CIT 4th floor)
			       

Interval Programming: A Multi-Objective Optimization Model for Motion
Planning in Autonomous Vehicles


			       Abstract

Interval Programming (IvP) is a new mathematical programming model for
use in multi-objective decision making problems. IvP places little or
no restrictions on objective function form while providing quick,
globally optimal solutions. Its development was motivated by the
behavior-based approach to implementing control of physical agents
navigating through a complex, changing environment. The action
selection problem of combining behaviors is best treated as a
multi-objective optimization problem, but solving such problems
accurately and quickly typically requires strict assumptions on
objective function form. Such assumptions can rarely be made in
behaviors that control vehicle navigation, which typically produce
objective functions that are multi-modal, discontinuous and defined
over a large action space precluding exhaustive evaluation or voting
methods. The IvP model addresses this issue by utilizing piecewise
linearly defined functions to represent the actual underlying
objective function that comes natural to a given behavior. We provide
an algorithm that searches through the combination space created by
pieces from each function. This algorithm is fast in practice and
guarantees that its solutions are globally optimal.


		  Host:  Professor Leslie Kaelbling