A unified end-to-end learning and control framework that is able to learn a (neural) control objective function, dynamics equation, control policy, or/and optimal trajectory in a control system.
https://wanxinjin.github.io/Pontryagin-Differentiable-Programming/