Development and evaluation of a robot-assisted gait rehabilitation system for post-stroke patients using real-time biofeedback and adaptive assist-as-needed control

Project Abstract

This study presents the design, implementation, and evaluation of a robot-assisted gait rehabilitation system for post-stroke patients that integrates real-time biofeedback with adaptive assist-as-needed (AAN) control to enhance motor recovery and functional walking ability. The system comprises a lower-limb exoskeleton integrated with a wireless motion capture module, electromyography sensors, and a multi-sensor feedback framework that provides patients with immediate, meaningful cues about gait parameters, trunk alignment, and muscle activation patterns. The AAN controller dynamically modulates robotic assistance based on real-time kinematic and kinetic error metrics, patient effort, and joint torque limits to promote active participation while ensuring safety and comfort. A two-phase study was conducted with 40 chronic stroke survivors randomized into an intervention group receiving robot-assisted gait training (RAGT) and a conventional treadmill training control group, over eight weeks (24 sessions). Primary outcomes included the 6-Minute Walk Test (6MWT), Stairs Negotiation Test, and Walking Index for Spinal Cord Injury II (WISCI II)-adapted measures, complemented by instrumented gait analysis to quantify symmetry, propulsion, and joint coordination (e.g., hip-knee-ankle coordination, spatiotemporal parameters). Secondary outcomes encompassed neuromuscular adaptations captured via surface EMG for tibialis anterior, gastrocnemius, and quadriceps, as well as metabolic cost (net oxygen consumption) and cardiovascular responses. Safety and tolerability were monitored through session adversity logs and vital signs. The results demonstrated that the RAGT group achieved statistically significant improvements in 6MWT distance (mean increase of 52.3 meters, p<0.01) and gait symmetry indices (single-support symmetry improved by 11.4%, p<0.05) compared with controls. EMG analyses revealed task-specific reorganization with increased plantarflexor engagement during push-off and improved tibialis anterior activation during swing, indicating better motor control and neuromechanical coupling. Biofeedback provided during training enhanced motor learning, as evidenced by reduced paretic limb activation delay and improved anticipatory muscle firing patterns. The AAN controller maintained safety by keeping peak knee torque within individualized thresholds (average reduction in peak knee moment by 8.7% without compromising task performance) and reduced compensatory strategies, such as hip hiking. Adherence and tolerability were high, with dropout rates below 5% and patient-reported comfort scores exceeding threshold levels across sessions. Subgroup analyses suggested greater benefits for individuals with moderate motor impairment and residual ankle dorsiflexion range, highlighting the potential for tailored rehabilitation dosing. The study also employed a mixed-methods approach, incorporating qualitative interviews to capture patient experiences with the device, perceived motivation, and perceived exertion, which aligned with quantitative outcomes and illuminated factors influencing engagement. In conclusion, the integrated real-time biofeedback and adaptive AA-N control in a robot-assisted gait rehabilitation system produced meaningful, multi-domain improvements in walking capacity, gait quality, and neuromuscular activation in post-stroke individuals, while maintaining safety and high user acceptability. These findings support the translational potential of combining adaptive robotics with biofeedback paradigms to optimize neurorehabilitation outcomes and inform clinical guidelines for implementing RAGT in standard care.

Project Overview

What This Project Is About

---

A straightforward, hands-on study of how a robot-assisted system can help people recover walking after a stroke. The project looks at using a robot to guide leg movements, while giving real-time feedback to the user and adjusting support automatically based on how much help the user needs.

---

---

The Problem It Addresses

---

Many stroke survivors have difficulty walking and rebuilding normal movement. Traditional therapy can be repetitive and time-consuming. This project explores a smarter way to assist, motivate, and tailor therapy to each person, potentially speeding up recovery and making rehab more engaging.

---

---

Objectives of the Project

---

1. Understand current gait rehabilitation limits and how robot assistance might help.
2. Develop a simple robot-assisted gait setup and a feedback system.
3. Implement adaptive assist-as-needed control that adjusts support in real time.
4. Test the system with healthy volunteers first, then with a small group of stroke models.
5. Evaluate safety, usability, and basic effectiveness signals.

---

---

What You Will Do Step by Step

---

1) Review basic rehab needs and existing devices. 2) Design a low-risk robot-assisted gait setup and feedback interface. 3) Implement a basic adaptive control that reduces or increases support as users improve. 4) Collect motion data during sessions with volunteers. 5) Analyze changes in movement quality and user feedback. 6) Discuss limitations and potential clinical steps.

---

---

Expected Outcome

Anticipated results include a working prototype, evidence that real-time feedback helps users stay engaged, and initial data showing safe, improved walking patterns under adaptive support. The project should suggest how this approach could fit into real rehab programs.