The paper is organised as follows. Section 2 begins with background material on MDPs, NMRDPs, and existing
approaches. Section 3 describes our new approach and
Section 4 presents NMRDPP.
Sections 5 and 6 report our
experimental analysis of the various approaches.
Section 7 explains how we used NMRDPP in the
competition. Section 8 concludes with remarks
about related and future work. Appendix B gives the
proofs of the theorems. Most of the material presented is compiled
from a series of recent conference and workshop papers
of the logic we use to represent rewards may be found in our 2005