Symbolic LAO* review

From: Bhushan Mandhani (bhushan_at_cs.washington.edu)
Date: Wed Nov 19 2003 - 10:06:27 PST

  • Next message: Daniel J. Klein: "Symbolic Heuristic Search for Factored Markov Decision Processes"

    Summary:
    The paper combines heuristic search with symbolic model checking
    (specifically the Spudd planner) for solving MDP's.

    Main Ideas:
    1. Symbolic model checking algorithms use dynamic programming to solve
    MDP's. They cover the entire state space, which can be large. Adding
    heuristic search, so that dynamic programming is done over a subset of
    states i.e, the states reachable from the start state, improves
    efficiency.

    2. In the LAO* search done here, everything is stored as an ADD. Further,
    the ADD's are made even more compact by "masking", and this further prunes
    the set of states considered.

    3. In the results in this paper, symbolic LAO* was seen to outperform
    Spudd.

    Flaw:
    1. The problems a1-a4 used in experimental evaluation are all artificial.
    The use of real-world MDP's to determine the gain in efficiency obtained
    using the proposed method would have been better.

    Future Work:
    Further exploration of general methods to prune the state space. The idea
    of combining heuristic search with dynamic programming seems powerful and
    should be explored further.


  • Next message: Daniel J. Klein: "Symbolic Heuristic Search for Factored Markov Decision Processes"

    This archive was generated by hypermail 2.1.6 : Wed Nov 19 2003 - 10:06:28 PST