<p>The Lagrange dynamics model can concisely describe the motion of complex multi-body systems while retaining their physical essence, providing a solid theoretical foundation for multi-robot formation control. On this basis, we propose a distributed position-attitude dynamic adaptation and formation control protocol suitable for Lagrangian multi-agent systems. This protocol achieves formation by integrating the positions of neighboring nodes to generate virtual node information. It incorporates the tanh function to construct a real-time dynamic adjustment mechanism that adjusts position for large deviations and yaw angle for small deviations. The algorithm is implemented under unknown dynamics and motion parameters, enhancing system robustness without requiring specific parameter customization, and is suitable for large-scale deployment. Numerical simulations demonstrate the effectiveness of the proposed method, and physical experiments confirm its stability in real-world scenarios.</p>