Comparative Benchmark of a Quantum Algorithm for the Bin Packing Problem

• Published in: 2022 IEEE Symposium Series on Computational Intelligence (SSCI) pp. 930 - 937
• Main Authors: De Andoin, Mikel Garcia, Oregi, Izaskun, Villar-Rodriguez, Esther, Osaba, Eneko, Sanz, Mikel
• Format: Conference Proceeding
• Language: English
• Published: IEEE 04.12.2022
• Subjects: Algorithms; Annealing; Benchmark testing; Bin Packing Problem; Combinatorial optimization; Heuristic algorithms; Measurement; Quantum algorithm; Quantum annealing; Quantum computation
• DOI: 10.1109/SSCI51031.2022.10022156

The Bin Packing Problem (BPP) stands out as a paradigmatic combinatorial optimization problem in logistics. Quantum and hybrid quantum-classical algorithms are expected to show an advantage over their classical counterparts in obtaining approximate solutions for optimization problems. We have recently proposed a hybrid approach to the one dimensional BPP in which a quantum annealing subroutine is employed to sample feasible solutions for single containers. From this reduced search space, a classical optimization subroutine can find the solution to the problem. With the aim of going a step further in the evaluation of our subroutine, in this paper we compare the performance of our procedure with other classical approaches. Concretely we test a random sampling and a random-walk-based heuristic. Employing a benchmark comprising 18 instances, we show that the quantum approach lacks the stagnation behaviour that slows down the classical algorithms. Based on this, we conclude that the quantum strategy can be employed jointly with the random walk to obtain a full sample of feasible solutions in fewer iterations. This work improves our intuition about the benefits of employing the scarce quantum resources to improve the results of a diminishingly efficient classical strategy.