A merge search algorithm and its application to the constrained pit problem in mining

Many large-scale combinatorial problems contain too many variables and constraints for conventional mixed-integer programming (MIP) solvers to manage. To make the problems easier for the solvers to handle, various meta-heuristic techniques can be applied to reduce the size of the search space, by removing, or aggregating, variables and constraints. A novel meta-heuristic technique is presented in this paper called merge search, which takes an initial solution and uses the information from a large population of neighbouring solutions to determine promising areas of the search space to focus on. The population is merged to produce a restricted sub-problem, with far fewer variables and constraints, which can then be solved by a MIP solver. Merge search is applied to a complex problem from open-pit mining called the constrained pit (CPIT) problem, and compared to current state-of-the-art results on well known benchmark problems minelib [7] and is shown to give better quality solutions in five of the six instances.