ABSTRACT

Existing ant colony algorithms only have one kind of pheromone. They use non-dominated solutions to update it while not making use of dominated solutions, which can provide valuable information for guiding the subsequent foraging process. To make full use of dominated solutions, we create a new kind of pheromone temporarily called a negative pheromone and propose a new ant colony optimisation algorithm called NMOACO/D, which combines MOEA/D-ACO with the negative pheromone. Many experiments have been carried out in this study to compare NMOACO/D with MOEA/D-ACO and other algorithms on several bi-objective travelling salesman problems. We demonstrate that NMOACO/D outperforms the MOEA/D-ACO and six different recently proposed related algorithms on all nine test instances. We also evaluate the effect of negative pheromone on the performance of the NMOACO/D. The results in this paper show that correctly making use of the information related to dominated solutions can further improve the ant colony algorithm performance.

摘要

现有的蚁群算法只有一种信息素。他们使用非支配解来更新它，而不使用支配解，这可以为指导后续觅食过程提供有价值的信息。为了充分利用支配解，我们创建了一种新的信息素，暂时称为负信息素，并提出了一种新的蚁群优化算法，称为 NMOACO/D，它将 MOEA/D-ACO 与负信息素相结合。本研究中进行了许多实验，将 NMOACO/D 与 MOEA/D-ACO 和其他算法在几个双目标旅行商问题上进行比较。我们证明 NMOACO/D 在所有九个测试实例上都优于 MOEA/D-ACO 和六种不同的最近提出的相关算法。我们还评估了负面信息素对 NMOACO/D 性能的影响。本文的结果表明，正确利用与支配解相关的信息可以进一步提高蚁群算法的性能。