Quantum assisted genetic algorithm for multi-objective optimization

Kurzfassung

Multi-objective optimization (MOO) remains a significant computational challenge, particularly for high-dimensional, non-convex problems which demand substantial computational effort from traditional evolutionary algorithms to obtain a Pareto front that is both well converged and uniformly diverse. To address this challenge, this work presents the quantum-assisted non-dominated sorting genetic algorithm (QANSGA), a novel hybrid heuristic that leverages principles of quantum annealing to enhance the core mechanisms of classical MOO genetic algorithm. The quantum component is realized via a D-Wave quantum annealer, which is used to implement quantum mutation operators and facilitate the injection of highly promising quantum solutions into the classical population pool. This hybrid mechanism fundamentally improves the population’s diversity maintenance, the global search capability and accelerates its convergence dynamics toward the true Pareto-optimal set, achieving the desired global optimum Pareto front in fewer generations. To rigorously evaluate its performance, we benchmark QANSGA against leading classical genetic algorithm (GA) for MOO problems, which is non-dominated sorting genetic algorithm (NSGA-II) across a widely used suite of complex multi-objective benchmark problems.

The results provide compelling evidence for the efficacy of quantum-inspired heuristics in overcoming the efficiency bottlenecks of conventional MOO methods. The study concluded that QANSGA outperforms classical NSGA-II in terms of convergence and diversity by 33% and 42% respectively on an average when applied to 6 benchmark multi-objective problems. This work positions QANSGA as a valuable new hybrid algorithm, paving the way for novel approaches to tackle multi-objective, complex optimization challenges in areas like energy system design, resource allocation, and logistics.