Researchers from Bio2Q and Keio University, along with industrial partners, have developed a new optimization framework that makes it much faster to solve large-scale combinatorial optimization problems by breaking them into smaller, independent pieces that can be solved in parallel.

Large optimization problems arise in applications such as vehicle routing, production scheduling, and facility planning, but they quickly become computationally expensive as their size and complexity increase. To address this, the researchers used quadratic unconstrained binary optimization (QUBO) to identify natural ways to partition a problem into weakly connected subproblems, allowing conventional optimization solvers to tackle each part simultaneously before combining the results. When applied to the capacitated vehicle routing problem, the approach reduced the number of optimization variables by up to 95%, reached in one minute the solution quality that conventional methods required 30 minutes to achieve, and successfully solved large problem instances that standard approaches could not complete within the allotted time.

By improving the scalability of existing optimization methods rather than replacing them, this work provides a practical framework for accelerating large-scale optimization across applications including logistics, scheduling, and facility planning.