A Control-Systems Approach to Understanding Human Learning

Our objective is to understand how humans control unknown dynamic systems. The internal model hypothesis (IMH), which is the predominant neuroscience theory of human learning, proposes that the brain constructs models of the body and its interactions with the physical world, and that these models are continuously updated and used for control. The proposed research applies engineering principles from dynamic systems and control theory to test the IMH and address other open human-learning questions. Specifically, this research addresses two main questions: (1) What control strategies do humans learn? (2) How do humans learn to control unknown dynamic systems? For (1), we propose a series of experiments designed to answer foundational questions on the strategies that humans use to control dynamic systems. The proposed experiments focus on: identifying the strategies that humans use for command following; understanding how humans interact with nonminimum-phase and high-relative-degree systems; characterizing the properties that make dynamic systems hard for humans to control; and identifying strategies humans use in interactive scenarios that involve more than one person controlling the same dynamic system (e.g., decentralized control). For (2), we propose to identify learning mechanisms that humans use to adapt to unknown dynamic systems. The proposed analyses include: exploring how humans use persistently ex- citing signals to learn; connecting human-learning mechanisms with indirect adaptive control theory; and understanding the mechanisms that humans use to avoid transient instability. This project offers a new paradigm in human learning research. Our system theoretic approach adopts analysis techniques from the fields of dynamic systems and control, and applies those techniques to answer fundamental questions of human learning. In addition, this project accomplishes interdisciplinary objectives. First, answering open questions in neuroscience also advances control system technology by identifying human learning mecha- nisms that are superior to existing control techniques. Second, improving our understanding of human learning also advances biomedical research by enabling new haptic, orthotic, and rehabilitation technologies.