A Review of Control Methods in Lower Limb Exoskeleton Robots: From Classical to Machine Learning Approaches

Lower limb exoskeleton robots have emerged as a transformative technology for rehabilitation, mobility assistance, and performance augmentation. These robotic systems are designed to support or restore lower limb movement in individuals with disabilities, injuries, or neurological disorders. The effectiveness of an exoskeleton largely depends on its control system, which governs movement, interaction with the user, and adaptation to environmental conditions. This paper explores the classification of lower limb exoskeleton robot controllers based on control strategies, input sources, loop types, and levels of autonomy. The study categorizes control strategies into predefined control, adaptive control, and supervisory control, each offering varying levels of responsiveness and adaptability. Additionally, different control input sources, such as biomechanical sensors, electromyography (EMG), and brain-machine interfaces (BMI), are analyzed for their role in detecting user intent. The paper also examines control loop types, differentiating between open-loop and closed-loop systems, and discusses how these impact stability and precision. Furthermore, the level of autonomy is explored, ranging from manual operation to fully autonomous systems that leverage artificial intelligence (AI) and machine learning. Despite advancements, challenges such as energy efficiency, sensor accuracy, real-time adaptability, and user comfort remain key limitations. This review synthesizes insights from existing literature into the state-of-the-art control strategies for lower-limb exoskeleton robots. Hence, it forms a roadmap for future advancements in the field.