A Surrogate-Assisted Continuous Ant Colony Optimization Method for Stochastic Optimization

Stochastic optimization problems, which involve random variables in the optimization process, are commonly seen in many applications such as engineering design and logistics management. The challenge of stochastic optimization is how to accurately evaluate the value of the objective function, while using a large number of Monte Carlo simulations for evaluation can result in significant computational costs. In this paper, we propose a surrogate-assisted continuous ant colony optimization method for stochastic optimization (SAACO). First, instead of using a lot of Monte Carlo simulations, SSACO estimates the approximated fitness values of the individuals in the population by a surrogate model. Second, the continuous ant colony optimization is used as the optimizer, where the population is evolved by constructing a Gaussian kernel function from global information to construct the promising solutions. The surrogate model is a forward model from the solution space to the objective space, while the Gaussian kernel function can be regarded as an inverse model from the objective space to the solution space. By combining two models, the performance of solving stochastic optimization problems can be improved. Experimental results demonstrate the promising performance of the proposed SAACO.