Evaluating Answer Set Clause Learning for General Game Playing

In games with imperfect information, the 'information set' is a collection of all possible game histories that are consistent with, or explain, a player's observations. Current game playing systems rely on these best guesses of the true, partially-observable game as the foundation of their decision making, yet finding these information sets is expensive. We apply reactive Answer Set Programming (ASP) to the problem of sampling information sets in the field of General Game Playing. Furthermore, we use this domain as a test bed for evaluating the effectiveness of oClingo, a reactive answer set solver, in avoiding redundant search by keeping learnt clauses during incremental solving.