Blending Grounding and Compilation for Efficient ASP Solving

Answer Set Programming (ASP) is a widely recognized formalism for Knowledge Representation and Reasoning. Traditional ASP systems, that employ the ground and solve architecture, are subject to the grounding bottleneck (i.e., variable-elimination can exhaust all computational resources). Compilation-based approaches have recently demonstrated how grounding can be effectively bypassed by compiling rules into propagators that simulate them. However, compiling an entire ASP program is not always advantageous. In this paper, we present both a program rewriting technique and an algorithm for the compilation of grounding that allow for unrestricted blending of grounding and compilation. We implement these techniques in a hybrid ASP system that compares favourably with state-of-the-art ASP solvers on established benchmarks.