基于跨邻域搜索的连续域蚁群优化算法

针对连续域蚁群算法寻优能力差、容易产生局部最优的问题,提出了一种基于跨邻域搜索的改进蚁群算法。首先,通过自适应种群划分方式计算可行解和不可行解群体;然后,针对不可行解群体利用自主选择学习算子选择对象进行学习,目的是不断扩大种群规模,避免算法陷入局部极值点,继而对可行解群体采取全局跨邻域搜索的方式,引导蚂蚁向全局最优解靠近,加快收敛速度;最后,基于全局最优解采用局部跨邻域的方式引导蚂蚁在小范围内进行细致搜索,提高收敛精度。通过与其他连续域蚁群优化算法针对CEC2017测试函数在低维和高维情况下的实验对比,证明本文算法具有较好的寻优能力和稳定性,能有效避免陷入局部最优。

Optimized ant colony algorithm for continuous domains based on across neighborhood search

To solve the problem that ant colony algorithm for continuous domain has poor search ability and tends to be trapped in local optima, an improved ant colony algorithm based on across neighborhood search is proposed. Firstly, feasible solution and unfeasible solution groups are selected by an adaptive population division method. Secondly, for the unfeasible solution group, a self-selected learning operator is used to choose the learning object in order to expand the population size constantly and avoid falling into local optima. Then global across neighborhood search is applied to the feasible solution group to guide the ants to the global optimal solution and accelerate the convergence . Finally, on the basis of global optimal solution, a local across neighborhood search method is used to guide the ants searching in a small range to improve the convergence accuracy. Experiments on the low-dimensional and high-dimensional cases of CEC2017 test functions reveal that, compared with other ant colony optimization algorithms for continuous domain, the proposed one has better optimization ability and stability, and can effectively escape from the local optimum.