Criteria-based parallelism for multiobjective problem solving

Computer scientists typically solve problems that require the optimization of some objective function. Given more than one objective to be simultaneously optimized, the problem becomes a Multicriteria Optimization Problem. Whereas the single objective version of a problem may be solved by use of well researched and studied algorithms, the Multicriteria version of the problem is typically not so quickly and easily solved. In fact, typically there is no polynomial-time algorithm for solving the multicriteria version, and the solution is not a single solution but rather a Pareto-Optimal set of solutions. In some cases it is possible to find an approximation of the solution set by solving Reduced Criteria Subproblems, then merging the solutions. This paper provides an overview of this approach, presents a parallel computing solution based on merging reduced-criteria subproblems, and gives consideration to the limitations of this methodology.