Learning-Based Model Predictive Control: Toward Safe Learning in Control

Abstract

Recent successes in the field of machine learning, as well as the availability of increased sensing and computational capabilities in modern control systems, have led to a growing interest in learning and data-driven control techniques. Model predictive control (MPC), as the prime methodology for constrained control, offers a significant opportunity to exploit the abundance of data in a reliable manner, particularly while taking safety constraints into account. This review aims at summarizing and categorizing previous research on learning-based MPC, i.e., the integration or combination of MPC with learning methods, for which we consider three main categories. Most of the research addresses learning for automatic improvement of the prediction model from recorded data. There is, however, also an increasing interest in techniques to infer the parameterization of the MPC controller, i.e., the cost and constraints, that lead to the best closed-loop performance. Finally, we discuss concepts that leverage MPC to augment learning-based controllers with constraint satisfaction properties.

Learning-Based Distributionally Robust Model Predictive Control of Markovian Switching Systems with Guaranteed Stability and Recursive Feasibility

“…Over the past decades, the decreasing cost of data acquisition, transmission and storage has caused a surge in research interest in data-driven approaches towards control. More recently, as the focus in research is gradually shifting towards real-life, safety-critical applications, there has been an increasing concern for safety guarantees of such datadriven methods, which are valid in a finite-data regime (see [1] for a recent survey).…”

Appropriate Smart Factory for SMEs: Concept, Application and Perspective

“…Modern decision-making technologies can contribute to resolving safety issues in automated manufacturing processes. These include tools for safety specification [49,50], risk-aware optimization [51,52], and safe learning [53]. In particular, a recent distributionally robust optimization technology may enhance safety in operation of autonomous systems when using inaccurate learning results [54].…”

Inverse Learning for Data-Driven Calibration of Model-Based Statistical Path Planning

“…Some recent research directions present interesting relations with the techniques proposed here and are worth some discussions. Recent reviews provide additional background on learning objectives from demonstrations [10], [11].…”