Abstract:
Abstract. A symbolic dynamic programming approach for solving first-order Markov decision processes within the situation calculus is presented. As an alternative specification language for dynamic worlds the fluent calculus is chosen and the fluent calculus formalization of the symbolic dynamic programming approach is provided. The major constructs of Markov decision processes such as the optimal value function and the policy are logically represented. The technique produces a set of first-order formulae that minimally partitions the state space. Consequently, the symbolic dynamic programming algorithm presented here does not require neither state nor action space enumeration, thereby solving the drawback of classical dynamic programming methods.
Citations
|
1928
|
Artificial Intelligence: A Modern Approach
– Russell, Norvig
- 2003
|
|
1717
|
Reinforcement Learning: An Introduction
– Sutton, Barto
- 1998
|
|
1424
|
Dynamic Programming
– Bellman
- 1957
|
|
524
|
The frame problem in the situation calculus: A simple solution (sometimes) and a completeness result for goal regression
– Reiter
- 1991
|
|
295
|
Planning and control
– Dean, Wellman
- 1991
|
|
168
|
Actions and Causal Laws
– McCarthy, Situations
- 1963
|
|
121
|
Decision theoretic planning: Structural assumptions and computational leverage
– Boutilier, Dean, et al.
- 1999
|
|
91
|
A new deductive approach to planning
– Hölldobler, Schneeberger
- 1990
|
|
77
|
Symbolic Dynamic Programming for FirstOrder MDPs
– Boutilier, Reiter, et al.
- 2001
|
|
29
|
A Formal Theory of
– FRENCH
- 1956
|
|
11
|
Decomposition principle for dynamic programs
– Dantzig, Wolfe
- 1960
|
|
6
|
The Fluent Calculus: A Specification Language for Robots with Sensors in Nondeterministic, Concurrent, and Ramifying Environments
– Thielscher
- 2000
|
|
4
|
GOLOG:A Logic Programming Language for Dynamic Domains
– Levesque, Lespérance
- 1997
|
|
1
|
Reinforcement Learning: A Survey. AI Access Foundation and
– Kaelbling, Moore
- 1996
|
