Abstract: To solve the problems associated with a large operator workload and complicated operations in the real-time control process of the flapping-wing flying robot, and to achieve distributed intelligent control of the robot, we propose a behavior planning method for the flapping-wing flying robot based on cluster analysis and motion description language (MDL). Using the results of a cluster analysis of the flying data of the flapping-wing flying robot, the motion behaviors of the robot are reasonably classified. While ensuring the relationship of the atoms in the MDL, the behavior characteristics of the flapping-wing flying robot are successfully extracted, and four motion atoms are defined in the winding rod task performed by the flapping-wing flying robot. This robot and an airborne gyroscope are then used to build an experimental robot system. Both physical and simulation experiments are conducted using the direct control method and the MDL-based robot behavior planning method, the results of which confirm the feasibility and effectiveness of the proposed method.