A Parallel Framework for Primal Heuristics in SHOT

The Supporting Hyperplane Optimization Toolkit (SHOT) is a solver designed for Mixed-Integer Nonlinear Programming (MINLP) problems. In its current form, SHOT handles primal heuristics sequentially on a single thread, which blocks the solver’s execution until the heuristic process completes. In this work, we utilize SHOT's modular architecture to develop a new approach that enables primal heuristics to run in parallel with the main optimization process, without blocking its progress. This allows for improved efficiency and scalability of the solver. The proposed framework will be evaluated on a representative set of benchmark problems from the MINLPLib library to assess its performance and effectiveness.