A neural network learns the optimal cost-to-go (value function) of an underactuated double pendulum. The trained network is then …