Closing the Loop Between Wearable Robots and Machine Learning: A New Paradigm for Steering Assistance Personalization Control

Lower-extremity wearable robotic devices, first introduced in the early 2000s, have been developed to enhance human mobility and support therapeutic training for patients. Recent advancements in human-in-the-loop (HIL) optimization have significantly improved the control of these devices, fine-tuning the interaction between humans and robots. This has led to more personalized assistance for daily living activities and rehabilitation training. Our comprehensive and extensive literature review, spanning from January 2017 to December 2023, highlights 34 noteworthy studies that have demonstrated enhanced human locomotion performance through HIL-optimized and personalized assistance. This review explores pivotal innovations and methodologies for controlling lower-extremity robotic exoskeletons, exosuits, and prostheses. It covers the establishment of control objectives, the application of various optimization methods, and the assessment of outcomes. Additionally, we provide a comparative analysis of the HIL optimization method against alternative control strategies, such as those based on reinforcement learning. Looking forward, we discuss expected trends that aim to enhance the efficacy of wearable robotic devices. We also recognize the challenges that need to be addressed to fully realize the benefits of customized gait assistance for individuals with lower-extremity impairments or neurological conditions. This includes technological, regulatory, and user-centered issues that could impact the widespread adoption and effectiveness of these innovative systems.