Core Boosting in SAT-Based Multi-objective Optimization

Maximum satisfiability (MaxSAT) constitutes today a successful approach to solving various real-world optimization problems through propositional encodings. Building on this success, approaches have recently been proposed for finding Pareto-optimal solutions to multi-objective MaxSAT (MO-MaxSAT) instances, i.e., propositional encodings under multiple objective functions. In this work, we propose core boosting as a reformulation/preprocessing technique for improving the runtime performance of MO-MaxSAT solvers. Core boosting in the multi-objective setting allows for shrinking the ranges of the multiple objectives at hand, which can be particularly beneficial for MO-MaxSAT relying on search that requires enforcing increasingly tighter objective bounds through propositional encodings. We show that core boosting is effective in improving the runtime performance of SAT-based MO-MaxSAT solvers typically with little overhead.