Abstract
Combinatorial optimization can be considered as the search in a discrete solution space. Genetic algorithm is one of multi-point search methods, and therefore, it has many advantages over one-point search methods. However, it has the following disadvantages: 1) Local optimization is not sufficient. 2) There exist many parameters and their tuning is annoying. To cope with these disadvantages, we propose a new multi-point search method using a lot of autonomous agents. Each agent executes local optimization, and all agents act cooperatively. As a result, sufficient local optimization, direct control of search process, and many advantages of multi-point search methods can be expected.
