Learning-Based Model Predictive Control for Safe Exploration (2018-03-22T00:00:00.000000Z)

TL;DR

This paper presents a learning-based model predictive control scheme that can provide provable high-probability safety guarantees and exploits regularity assumptions on the dynamics in terms of a Gaussian process prior to construct provably accurate confidence intervals on predicted trajectories.

Abstract

Learning-based methods have been successful in solving complex control tasks without significant prior knowledge about the system. However, these methods typically do not provide any safety guarantees, which prevents their use in safety-critical, real-world applications. In this paper, we present a learning-based model predictive control scheme that can provide provable high-probability safety guarantees. To this end, we exploit regularity assumptions on the dynamics in terms of a Gaussian process prior to construct provably accurate confidence intervals on predicted trajectories. Unlike previous approaches, we do not assume that model uncertainties are independent. Based on these predictions, we guarantee that trajectories satisfy safety constraints. Moreover, we use a terminal set constraint to recursively guarantee the existence of safe control actions at every iteration. In our experiments, we show that the resulting algorithm can be used to safely and efficiently explore and learn about dynamic systems.

Authors

A. Krause

4 papers

Felix Berkenkamp

4 papers

Torsten Koller

2 papers

TL;DR

Abstract

Authors

References30 items

Learning and Control Using Gaussian Processes

Linear Model Predictive Safety Certification for Learning-Based Control

Cautious NMPC with Gaussian Process Dynamics for Autonomous Miniature Race Cars

Ellipsoidal Estimates of Reachable Sets for Control Systems with Nonlinear Terms

Safe Model-based Reinforcement Learning with Stability Guarantees

Robust Constrained Learning-based NMPC enabling reliable mobile robot path tracking

Gaussian Process Model Predictive Control of an Unmanned Quadrotor

Safe learning of regions of attraction for uncertain, nonlinear systems with Gaussian processes

A provably correct MPC approach to safety control of urban traffic networks

Model-based reinforcement learning with parametrized physical models and optimism-driven exploration

Reachability-based safe learning with Gaussian processes

A General-Purpose Software Framework for Dynamic Optimization (Een algemene softwareomgeving voor dynamische optimalisatie)

A robust model predictive control algorithm augmented with a reactive safety mode

Provably safe and robust learning-based model predictive control

PILCO: A Model-Based and Data-Efficient Approach to Policy Search

Reinforcement Learning Versus Model Predictive Control: A Comparison on a Power System Problem

Approximation of convex sets by polytopes

Robust variable horizon model predictive control for vehicle maneuvering

On the implementation of an interior-point filter line-search algorithm for large-scale nonlinear programming

Gaussian process model based predictive control

Gaussian Processes For Machine Learning

Closed-loop stochastic dynamic process optimization under input and state constraints

A deterministic algorithm for global optimization

Spline Models for Observational Data

Additions

Ellipsoidal Calculus For Estimation And Control

Model Predictive Control: Theory and Design

Control of Uncertain Nonlinear Systems Using Ellipsoidal Reachability Calculus

Gaussian process optimization in the bandit setting: No regret and experimental design

Reinforcement Learning: An Introduction

Field of Study

Journal Information

Name

Page

Venue Information

Name

Type

URL

Alternate Names