Control Algorithms for Medical Recovery Robots Used in Physiotherapy

Rehabilitation robots are an essential component of modern physical therapies, enabling personalized and in-tensive exercises for patients with neuromotor impairments. This paper provides a structured review of the main categories of control algorithms used in robotic therapy systems, with a focus on their ability to adapt to the patient’s condition and variability during exercises. Adaptive control strategies, robust and predictive control, as well as intelligent algorithms based on machine learning and deep neural networks are analyzed. Special attention is paid to hybrid control schemes, as well as techniques that integrate surface electromyography (sEMG) signals and virtual constraints (Virtual Fixtures). Each control category is discussed in light of relevant recent works, highlighting the advantages, implementation challenges and clinical implications. The comparative analysis highlights the upward trend towards hybrid and data-driven methods, which allow for real-time personalization, increased robustness and improved functional outcomes. The paper concludes with a presentation of the main research challenges, such as real-time adaptation to nonlinear dynamics, safety in human–robot interaction, and interpretability of learning-based control strategies in a clinical context.