Algorithms of data generation for deep learning and feedback design: A survey

Recent research reveals that deep learning is an effective way of solving high dimensional Hamilton-Jacobi-Bellman equations. The resulting feedback control law in the form a neural network computationally efficient for real-time applications optimal control. A critical part this design method to generate data training and validating its accuracy. In paper, we provide survey existing algorithms can be used data. All surveyed paper are causality-free, i.e., solution at point computed without using value function any other points. At end illustrative example given.