Abstract: A strategy tree concept is proposed to address the evolving needs of autonomous adaptability of systems in complex and strong adversarial environments. A system strategy tree is designed that covers the adjustment branches of perception resources, processing resources, decision-making resources, and strike resources. An adaptive sample simulation generation technology based on a trial experiment is proposed to solve the problem of limited quantities of adaptive sample data. A method is proposed for discrete clustering and mining association rules using continuous adaptive conditional attributes. Based on the frequent term mining of adaptive samples, a new branch of the strategy tree is generated to achieve iterative evolution of the autonomous adaptive ability of systems and solve the problem that traditional rule systems cannot automatically discover new knowledge. A validation environment with 50 nodes is built to verify the application effectiveness of strategy trees with different generations. The simulation test results show that the application effect of the strategy tree after iterative evolution is significantly better than that of the initial or previous generation strategy tree.