In recent years, UAVs have been widely used in a variety of fields such as firefighting, emergency communications, emergency lighting and surveillance. However, how to ensure the stable landing of UAVs on mobile platforms is still an issue of concern. Therefore, for the multimodal stability control challenges of autonomous landing of UAVs in complex dynamic environments, a cooperative control algorithm based on dynamic weight Proximal Policy Optimization (PPO) is proposed, the core of which is to reconstruct the positional error reward function through a dynamic weight adjustment mechanism and to establish a reinforcement learning model containing initial state constraints. The method addresses the conflict between trajectory tracking accuracy and attitude adjustment efficiency through the dynamic reward weight allocation strategy for stationary and moving dual-state landing platform scenarios. Simulation results show that the method can adapt to different platform motion noise conditions, providing a feasible technical path for robust autonomous landing control of UAVs in complex dynamic environments, which is of methodological reference value for multi-objective co-optimization research in the field of motion control of intelligent bodies.