Heterogeneous multi-unmanned systems enable operations in complex environments and diverse tasks but face challenges due to differing dynamics and environmental models. To address this, we propose a hierarchical control framework with homogeneous sub-formations and heterogeneous coordination between formations. A SAC-based cooperative path planning method is developed, combining Jump Point Search (JPS) for global path optimization with Soft Actor-Critic (SAC) for local adjustments by leaders. Followers use SAC for formation keeping and obstacle avoidance. Experiments show a planning success rate of 99.3 \(\%\) in homogeneous and 95.3 \(\%\) in heterogeneous scenarios, demonstrating the method’s feasibility and robustness.