Abstract: Focusing on the design of land battlefield information system, combined with the latest achievements of distributed computing,based on the concept of "cloud-network-end" fusion, a set of information system architecture suitable for distributed battlefield nodes and intelligent requirements is constructed. Aiming at the key requirements of intelligent application and enhancement in the constructed architecture,such characteristics as constrained distributed land battlefield node resources, diversified tasks, etc are synthesized, and on the basis of hierarchical federal learning, self-adaptive adjustment is done according to the status of resources. The operation means for keeping the balance between the resource expenditure and intelligence improvement are provided. It has been verified by experiments that the proposed architecture and method have better results, and can better adapt to the intelligence requirements of land battlefield nodes under distributed conditions.