Constrained model-free reinforcement learning for process optimization

Abstract

Reinforcement learning (RL) is a control approach that can handle nonlinear stochastic optimal control problems. However, despite the promise exhibited, RL has yet to see marked translation to industrial practice primarily due to its inability to satisfy state constraints. In this work we aim to address this challenge. We propose an “oracle”-assisted constrained Q-learning algorithm that guarantees the satisfaction of joint chance constraints with a high probability, which is crucial for safety critical tasks. To achieve this, constraint tightening (backoffs) are introduced and adjusted using Broyden’s method, hence making the backoffs self-tuned. This results in a methodology that can be imbued into RL algorithms to ensure constraint satisfaction. We analyze the performance of the proposed approach and compare against nonlinear model predictive control (NMPC). The favorable performance of this algorithm signifies a step towards the incorporation of RL into real world optimization and control of engineering systems, where constraints are essential.