An Embedded Feature Selection Framework for Control

TL;DR

This paper introduces D-AFS, a dual-world embedded feature selection framework for control systems that efficiently identifies relevant sensors, improving control performance and reducing sensor requirements in dynamic environments.

Contribution

The novel D-AFS framework leverages dual-world analysis in DRL to select optimal sensors, addressing the limitations of traditional feature selection in dynamic control domains.

Findings

D-AFS achieved 18.7% drag reduction with 5 probes.

D-AFS outperformed state-of-the-art solutions and human experts.

Successfully applied to multiple classical control problems.

Abstract

Reducing sensor requirements while keeping optimal control performance is crucial to many industrial control applications to achieve robust, low-cost, and computation-efficient controllers. However, existing feature selection solutions for the typical machine learning domain can hardly be applied in the domain of control with changing dynamics. In this paper, a novel framework, namely the Dual-world embedded Attentive Feature Selection (D-AFS), can efficiently select the most relevant sensors for the system under dynamic control. Rather than the one world used in most Deep Reinforcement Learning (DRL) algorithms, D-AFS has both the real world and its virtual peer with twisted features. By analyzing the DRL's response in two worlds, D-AFS can quantitatively identify respective features' importance towards control. A well-known active flow control problem, cylinder drag reduction, is used for evaluation. Results show that D-AFS successfully finds an optimized five-probes layout with 18.7\% drag reduction than the state-of-the-art solution with 151 probes and 49.2\% reduction than five-probes layout by human experts. We also apply this solution to four OpenAI classical control cases. In all cases, D-AFS achieves the same or better sensor configurations than originally provided solutions. Results highlight, we argued, a new way to achieve efficient and optimal sensor designs for experimental or industrial systems. Our source codes are made publicly available at https://github.com/G-AILab/DAFSFluid.