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

TLDR

Reinforcement learning excels at solving complex tasks in unknown environments but typically offers no safety guarantees, limiting its use in safety‑critical real‑world applications. This paper proposes a learning‑based model predictive control framework that ensures high‑probability safety throughout the learning process. The approach builds a reliable statistical model to generate confidence‑interval‑based trajectory predictions, incorporates input‑dependent uncertainties, enforces safety constraints, and uses a terminal set to recursively guarantee safe control actions. Experiments demonstrate that the algorithm can safely explore an inverted pendulum and solve a cart‑pole reinforcement learning task while respecting safety constraints.

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.