UNIVERSITY OF HERTFORDSHIRE
		COMPUTER SCIENCE RESEARCH COLLOQUIUM

			     presents


                 "Planning for Human-Robot Interaction: 
                   Representing Time and Intention"

		           Dr. Frank Broz 
	          (Adaptive Systems Research Group,
		    University of Hertfordshire)


		      18 February 2009 (Wednesday)

			 Lecture Theatre B404 
		   Hatfield, College Lane Campus
			      3 - 4 pm

	    Coffee/tea and biscuits will be available.

		   Everyone is Welcome to Attend



Abstract:

 This research concerns planning for a specific class of human-robot 
interaction domains: those in which robots engage in tasks with humans 
that are governed by social conventions. Recognizing the intentions 
of others and performing actions at the appropriate time in the 
interaction are critical abilities for social competence. These 
interactions can be represented by probabilistic graphical models, 
and policies for the robot's role can be solved for using decision 
theoretic planning techniques. Because the human participant is part 
of the robot's environment, the world state in these problems is both 
dynamic and partially observable. In this work, human intention is 
represented as hidden state in a partially observable Markov decision 
process (POMDP), and the time-dependence of action outcomes are 
explicitly modeled using a time-indexed state space. A novel state 
aggregation method that acts over the time dimension of the state space 
is designed to trade off between the accuracy of the representation 
and its size in order to find sufficiently expressive models that 
can also be solved tractably. The utility of this approach is 
demonstrated by implementing a controller for a car that performs a 
social task with human drivers in a driving simulator. Performance 
is evaluated by comparing the policies obtained using the proposed 
modeling technique to policies developed using less expressive 
representations. In interactions with human participants, the 
policies for time-dependent POMDP models with human intention as 
hidden state outperform the other policies, achieving both higher 
rewards and more positive evaluations for naturalness and social 
propriety of behavior.

---------------------------------------------------
Hertfordshire Computer Science Research Colloquium
http://homepages.feis.herts.ac.uk/~nehaniv/colloq