Adaptive Cyber‐Physical‐Human Systems: Exploiting Cognitive Modeling and Machine Learning in the Control Loop

TLDR

Cyber‑physical‑human systems integrate sensors, computers, communication devices, and humans to accomplish mission tasks, allowing dynamic connectivity and varied human roles; adaptive variants rely on mutual adaptation driven by prior knowledge, cognitive modeling, and online machine learning to respond to changing contexts. The study identifies key challenges to realizing adaptive cyber‑physical‑human systems. The authors propose a functional reference architecture that delineates learning and adaptation processes, human and CPS roles, and guides development of adaptive CPH systems.

Abstract

ABSTRACT Cyber‐physical‐human (CPH) systems are purposeful arrangements of sensors, computers, communication devices, and humans to perform tasks that achieve specific mission objectives. These systems typically allow other systems, devices, and data streams to connect/disconnect as needed during mission execution. The roles of humans in CPH systems are quite varied. In adaptive CPH systems, humans collaborate with the cyber‐physical elements to jointly accomplish tasks, and adapt to changing contexts to accomplish mission goals. Mutual adaptation based on prior knowledge, cognitive modeling, and online machine learning are key characteristics of adaptive CPH systems. This paper presents key challenges in realizing adaptive CPH systems. It discusses learning and adaptation, as well as human and CPS roles in adaptive CPH systems. It offers a functional (reference) architecture to inform and guide the development of adaptive CPH systems. It concludes with a discussion of research needed to advance the state‐of‐the‐art of adaptive CPH systems.