Abstract: Multi-MA（mobile agents） cooperative systems lead to some flexible and efficient solutions to distributed systems. Some current research works are limited in the dynamic network environment. Based on the model of a Multi-MA cooperative system, the specific features of knowledge of agents in dynamic environment are taken into account, such as uncertainty, influence, and possible contradiction between agents. And in this paper, some notions, a theoretic model and an algorithm on fuzzy knowledge representation and exchange are proposed. The theory enables the key issues to be well solved, including fuzzy belief composition and contradiction coordination. With the help of some detail simulation experiments, what the authors proposed is applied to dynamic network topology discovery. As expected, the model is proved to be feasible and efficient.