Learning-based Model Predictive Control for Safe Exploration and Reinforcement Learning

TL;DR

This paper introduces a learning-based model predictive control method that ensures safety guarantees during reinforcement learning in unknown environments, enabling safe exploration in real-world applications.

Contribution

It presents a novel MPC scheme with probabilistic safety guarantees using input-dependent uncertainty modeling and recursive safety constraints.

Findings

Successfully applied to inverted pendulum for safe exploration.

Demonstrated safe reinforcement learning on cart-pole system.

Provided high-probability safety guarantees throughout learning process.

Abstract

Reinforcement learning has been successfully used to solve difficult tasks in complex unknown environments. However, these methods typically do not provide any safety guarantees during the learning process. This is particularly problematic, since reinforcement learning agent actively explore their environment. This prevents their use in safety-critical, real-world applications. In this paper, we present a learning-based model predictive control scheme that provides high-probability safety guarantees throughout the learning process. Based on a reliable statistical model, we construct provably accurate confidence intervals on predicted trajectories. Unlike previous approaches, we allow for input-dependent uncertainties. Based on these reliable 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. We evaluate the resulting algorithm to safely explore the dynamics of an inverted pendulum and to solve a reinforcement learning task on a cart-pole system with safety constraints.