Learned Models for Continuous Planning

Authors:

Matthew D. Schmill
Experimental Knowledge Systems Laboratory
University of Massachusetts
Box 34610
Amherst, MA 01003-4610
E-mail: schmill@cs.umass.edu
Phone: (413) 545-3616
Fax: (413) 545-1249

Tim Oates
Experimental Knowledge Systems Laboratory
University of Massachusetts
Box 34610
Amherst, MA 01003-4610
E-mail: oates@cs.umass.edu
Phone: (413) 577-0669
Fax: (413) 545-1249

Paul R. Cohen
Experimental Knowledge Systems Laboratory
University of Massachusetts
Box 34610
Amherst, MA 01003-4610
E-mail: cohen@cs.umass.edu
Phone: (413) 545-3638
Fax: (413) 545-1249

Abstract:

We are interested in the nature of activity -- structured behavior of nontrivial duration -- in intelligent agents. We believe that the development of activity is a continual process in which simpler activities are composed, via planning, to form more sophisticated ones in a hierarchical fashion. The success or failure of a planner depends on its models of the environment, and its ability to implement its plans in the world. We describe an approach to generating dynamical models of activity from real-world experiences and explain how they can be applied towards planning in a continuous state space.

Keywords:

learning, time series analysis, dynamic time warping, clustering, piecewise regression, robotics, planning

Availability:

EKSL Publications Page