Techniques for efficient lazy-grounding ASP solving

Answer-Set Programming (ASP) is a well-known and expressive logic programming paradigm based on efficient solvers. State-of-the-art ASP solvers require the ASP program to be variable-free, they thus ground the program upfront at the cost of a potential exponential explosion of the space required. Lazy-grounding, where solving and grounding are interleaved, circumvents this grounding bottleneck, but the resulting solvers lack many important search techniques and optimizations. The recently introduced ASP solver Alpha combines lazy-grounding with conflict-driven nogood learning (CDNL), a core technique of efficient ASP solving. This work presents how techniques for efficient propagation can be lifted to the lazy-grounding setting. The Alpha solver and its components are presented and detailed benchmarks comparing Alpha to other ASP solvers demonstrate the feasibility of this approach.