Data Development and Deep Learning for HJB Equations

Wei Kang
Naval Postgraduate School

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 of a neural network is computationally efficient for real-time applications of optimal control. A critical part of this design method is to generate data for training the neural network and validating its accuracy. This talk is a survey of existing algorithms that can be used to generate data. All the algorithms surveyed here are causality-free, i.e., the solution at a point is computed without using the value of the function at any other points. Examples will be presented to illustrate the method with application to optimal feedback design using solutions learned from the data that is generated for the associated HJB equations.

Presentation (PDF File)

Back to Workshop I: High Dimensional Hamilton-Jacobi Methods in Control and Differential Games