In this paper, we use Deep Lagrangian Networks (DeLaN) that combine Lagrangian mechanics and deep learning to model the rigid-body dynamics. In addition, we implemente the model using the Autograd function of the deep learning library, which allows us to build more complex neural networks in the dynamics model. We also evaluated the activation function of the neural networks and selected the most appropriate activation function to improve the model performance. Finally, we conduct experiments

more »

... th a planar double pendulum model, and the experimental results proved the effectiveness of our methods.