On Coalition Generation Algorithm in MAS Control System

The multi-agent cooperation is an important task in a distributed control system. Its key part is how to generate coalition group of agents. The ant algorithm is used often to solve this problem in MAS control system. But it has some deficiencies, such as partial optimization, slow constringency and low precision. With the respective virtues of genetic algorithm and ant algorithm, a new ant algorithm was presented in this work. It included two phase. In the first phase, it produced the information elements through genetic algorithm, and obtained optimization coalition through ant algorithm in the second phase. The simulation experiments were done and the results show that both the optimizing performance and time performance of new ant algorithm are improved enormously. The new algorithm not only quickens the constringency process and advances the precision of calculation result, but also avoids partial optimization problem.