A fast and robust algorithm for solving biobjective mixed integer programs

We present a fast and robust algorithm for solving biobjective mixed integer linear programs. Two existing methods are studied: \(\epsilon \) $ϵ$ -Tabu Method and the Boxed Line Method. By observing structural characteristics of nondominated frontiers and computational bottlenecks, we develop enhanced versions of each method. Limitations of the current state of test instances are observed, and a new body of instances are generated to diversify computational standards. We demonstrate efficacy with a computational study. The enhancement to \(\epsilon \) $ϵ$ -Tabu Method offers an average speed-up factor of 3 on some instances; the enhancement to Boxed Line Method offers an average speed-up factor of 18 on some instances. A hybrid, two-phase method is designed to leverage the strengths of each method with its corresponding enhancement, thus having a robust approach to a wider range of instances; it outperforms on all instances with a typical speed-up factor of 2–3. We also demonstrate that it is capable of producing a high-quality approximation of the nondominated frontier in a fraction of the time required to produce the complete nondominated frontier.