Chapter

Advances in Knowledge Representation, Logic Programming, and Abstract Argumentation

Volume 9060 of the series Lecture Notes in Computer Science pp 33-46

Simulation of Action Theories and an Application to General Game-Playing Robots

  • Michael ThielscherAffiliated withSchool of Computer Science and Engineering, The University of New South Wales

* Final gross prices may vary according to local VAT.

Get Access

Abstract

We consider the problem of verifying whether one action theory can simulate a second one. Action theories provide modular descriptions of state machines, and simulation means that all possible sequences of actions in one transition system can be matched by the other. We show how Answer Set Programming can be used to automatically prove simulation by induction from an axiomatisation of two action theories and a projection function between them. Our interest in simulation of action theories comes from general game-playing robots as systems that can understand the rules of new games and learn to play them effectively in a physical environment. A crucial property of such games is their playability, that is, each legal play sequence in the abstract game must be executable in the real environment.