Automated Reasoning in the Simulation of Evolvable Systems

We present a novel application of automated theorem proving for the logical simulation of evolvable systems. Modelled using a logical framework, these systems are built hierarchically from components where each component is specified as a first order theory and may have an associated supervisory component. The supervisory component monitors and possibly changes its associated component. The simulation of this framework makes intensive use of automated theory proving – when running a simulation, almost all computational steps are those of a theorem prover. We present this novel combination of a logical setting involving meta-level logics and large sets of formulae for system description, together with theorem proving requirements which involve often slowly changing specifications with the need for rapid assessment of deducibility and consistency. We illustrate how theorem provers are used using an evolvable extension of the blocks world and present a caching structure to reduce simulation times. We then evaluate the suitability of several theorem provers for this application.