Neural network architectures using min-plus algebra for solving certain high dimensional optimal control problems and Hamilton-Jacobi PDEs

Abstract

Solving high dimensional optimal control problems and corresponding Hamilton-Jacobi PDEs are important but challenging problems in control engineering. In this paper, we propose two abstract neural network architectures which are respectively used to compute the value function and the optimal control for certain class of high dimensional optimal control problems. We provide the mathematical analysis for the two abstract architectures. We also show several numerical results computed using the deep neural network implementations of these abstract architectures. A preliminary implementation of our proposed neural network architecture on FPGAs shows promising speed up compared to CPUs. This work paves the way to leverage efficient dedicated hardware designed for neural networks to solve high dimensional optimal control problems and Hamilton-Jacobi PDEs.

References in corpus (27)

• Artificial Neural Networks for Solving Ordinary and Partial Differential Equations
• DGM: A deep learning algorithm for solving partial differential equations
• Solving high-dimensional partial differential equations using deep learning
• fPINNs: Fractional Physics-Informed Neural Networks
• Deep learning-based numerical methods for high-dimensional parabolic partial differential equations and backward stochastic differential equations
• A composite neural network that learns from multi-fidelity data: Application to function approximation and inverse PDE problems
• A unified deep artificial neural network approach to partial differential equations in complex geometries
• PPINN: Parareal Physics-Informed Neural Network for time-dependent PDEs
• PDE-Net 2.0: Learning PDEs from Data with A Numeric-Symbolic Hybrid Deep Network
• Quantifying total uncertainty in physics-informed neural networks for solving forward and inverse stochastic problems
• Adversarial Uncertainty Quantification in Physics-Informed Neural Networks
• Machine learning approximation algorithms for high-dimensional fully nonlinear partial differential equations and second-order backward stochastic differential equations
• Solving Many-Electron Schrödinger Equation Using Deep Neural Networks
• Deep learning observables in computational fluid dynamics
• Functional Multi-Layer Perceptron: a Nonlinear Tool for Functional Data Analysis
• The max-plus finite element method for solving deterministic optimal control problems: basic properties and convergence analysis
• Polynomial approximation of high-dimensional Hamilton-Jacobi-Bellman equations and applications to feedback control of semilinear parabolic PDEs
• A Mean-Field Optimal Control Formulation of Deep Learning
• On some neural network architectures that can represent viscosity solutions of certain high dimensional Hamilton--Jacobi partial differential equations
• Reachability-Based Safety and Goal Satisfaction of Unmanned Aerial Platoons on Air Highways
• Curse of dimensionality reduction in max-plus based approximation methods: theoretical estimates and improved pruning algorithms
• Perspectives on characteristics based curse-of-dimensionality-free numerical approaches for solving Hamilton-Jacobi equations
• QRnet: optimal regulator design with LQR-augmented neural networks
• A new scalable algorithm for computational optimal control under uncertainty
• A max-plus dual space fundamental solution for a class of operator differential Riccati equations
• On Decomposition Models in Imaging Sciences and Multi-time Hamilton-Jacobi Partial Differential Equations
• A Deterministic Affine-Quadratic Optimal Control Problem