ABSTRACT. The grounding bottleneck is an important open issue in Answer Set Programming. Lazy grounding addresses it by interleaving grounding and search. The performance of current lazy-grounding solvers is not yet comparable to that of ground-and-solve systems, however. The aim of this thesis is to extend prior work on lazy grounding by novel heuristics and other techniques like non-ground conflict learning in order to speed up solving. Parts of expected results will be beneficial for ground-and-solve systems as well.