Learning an Approximate Model Predictive Controller with Guarantees

Abstract

A supervised learning framework is proposed to approximate a model predictive controller (MPC) with reduced computational complexity and guarantees on stability and constraint satisfaction. The framework can be used for a wide class of nonlinear systems. Any standard supervised learning technique (e.g. neural networks) can be employed to approximate the MPC from samples. In order to obtain closed-loop guarantees for the learned MPC, a robust MPC design is combined with statistical learning bounds. The MPC design ensures robustness to inaccurate inputs within given bounds, and Hoeffding's Inequality is used to validate that the learned MPC satisfies these bounds with high confidence. The result is a closed-loop statistical guarantee on stability and constraint satisfaction for the learned MPC. The proposed learning-based MPC framework is illustrated on a nonlinear benchmark problem, for which we learn a neural network controller with guarantees.

6 pages, 3 figures, to appear in IEEE Control Systems Letters

References in corpus (2)

• A Brief Survey of Deep Reinforcement Learning
• Learning an Approximate Model Predictive Controller with Guarantees

Cited by in corpus (25)

• Efficient representation and approximation of model predictive control laws via deep learning
• Learning an Approximate Model Predictive Controller with Guarantees
• Safe and Fast Tracking on a Robot Manipulator: Robust MPC and Neural Network Control
• Robust Tube-based Model Predictive Control with Koopman Operators--Extended Version
• Physics-Informed Neural Nets for Control of Dynamical Systems
• Deep Learning-Based Model Predictive Control for Resonant Power Converters
• On Recurrent Neural Networks for learning-based control: recent results and ideas for future developments
• Robust Learning-based Predictive Control for Discrete-time Nonlinear Systems with Unknown Dynamics and State Constraints
• Probabilistic performance validation of deep learning-based robust NMPC controllers
• Reliably-stabilizing piecewise-affine neural network controllers
• Data-driven computation of invariant sets of discrete time-invariant black-box systems
• Online learning with stability guarantees: A memory-based real-time model predictive controller
• Deep Model Predictive Control with Stability Guarantees
• Neural Lyapunov Differentiable Predictive Control
• Combining Federated Learning and Control: A Survey
• Approximate non-linear model predictive control with safety-augmented neural networks
• Performance guaranteed MPC Policy Approximation via Cost Guided Learning
• Constraint-adaptive MPC for large-scale systems: Satisfying state constraints without imposing them
• Dynamic Controlled Variables Based Dynamic Self-Optimizing Control
• Imitation Learning of MPC with Neural Networks: Error Guarantees and Sparsification
• An Improved Data Augmentation Scheme for Model Predictive Control Policy Approximation
• Polytopic Input Constraints in Learning-Based Optimal Control Using Neural Networks
• A Sensitivity-based Data Augmentation Framework for Model Predictive Control Policy Approximation
• A Neural Network-based Multi-timestep Command Governor for Nonlinear Systems with Constraints
• YANNs: Y-wise Affine Neural Networks for Exact and Efficient Representations of Piecewise Linear Functions