Smart Assistive Exoskeleton for Gait Rehabilitation Using Real-Time EMG-Driven Control

Project Abstract

This study presents the development and evaluation of a smart assistive exoskeleton designed to enhance gait rehabilitation through real-time electromyography (EMG)-driven control. The system integrates lightweight, modular actuators with a multi-sensor fusion framework to deliver synchronized ankle, knee, and hip assistance aligned with the user’s residual motor intent. The core innovation lies in a real-time EMG processing pipeline that translates muscle activation patterns into adaptive torque commands using a hybrid control strategy that couples supervised machine learning models with a robust model-based controller. EMG signals from major lower-limb muscle groups are preprocessed to mitigate noise and motion artifacts, then mapped to a personalized assist-as-needed (AaN) profile that adjusts impedance and timing in response to the user’s gait phase and speed. A kinematic/kinetic estimation module leverages inertial measurement units (IMUs) and servo-torque sensing to estimate joint angles, limb trajectories, and interaction forces, enabling closed-loop adaptation and ensuring safe joint limits and ergonomic joint trajectories. The exoskeleton’s control framework comprises three layers intent detection, reference trajectory generation, and torque synthesis. Intent detection uses a combination of pattern recognition and time-series classification to identify the user’s walking intention, including initiation, cadence, and propulsion demands. Reference trajectories are computed from a subject-specific gait database and online refinements, ensuring smooth transitions between stance and swing phases. Torque synthesis employs a feedforward-feedforward-with-feedback strategy that combines a model-based impedance controller with a data-driven residual torque predictor to compensate for parasitic dynamics and inter-limb coupling. The system prioritizes minimal latency, achieving sub-40 ms response times to maintain naturalistic movement and reduce cognitive load. A cohort of twenty able-bodied and twenty individuals with chronic stroke participated in a multi-phase evaluation, including treadmill and overground walking tests at multiple speeds. Primary outcomes included gait speed, step length symmetry, toe clearance, and energy expenditure (net metabolic cost). Secondary outcomes encompassed in-shoe pressure distribution, joint kinematic accuracy, user workload (NASA-TLX), and functional mobility (Timed Up and Go, 10-Meter Walk Test). Results demonstrate statistically significant improvements in gait symmetry and speed for both cohorts when using the EMG-driven exoskeleton compared to baseline walking and conventional passive exoskeletons. Notably, stroke participants exhibited enhanced propulsion symmetry and reduced compensatory trunk movements, translating to lower metabolic costs and improved stability during task demands. Real-time adaptation maintained comfortable interaction forces (within acceptable ranges) throughout varying terrains and speeds, with minimal calibration time required per user. The findings validate the feasibility and effectiveness of EMG-driven control for intelligent assistive devices in neurorehabilitation, highlighting the potential to personalize therapy, accelerate motor recovery, and facilitate autonomous community ambulation. Limitations include variability in EMG signal quality across users and potential fatigue effects, which are addressed through adaptive filtering, online recalibration, and fatigue-aware control policies. Future work will explore scalable online learning for broad patient populations, integration with virtual reality–augmented therapy, and long-term clinical trials to assess retention and transfer of gains to real-world activities.

Project Overview

What This Project Is About

---

A straightforward exploration of how a wearable exoskeleton can help people regain walking ability by using muscle signals to guide assistance in real time.

---

---

The Problem It Addresses

---

---

---

Objectives of the Project

---

1. Understand the basics of gait and assistive devices.
2. Develop a simple EMG-based control concept for providing leg assistance.
3. Design a lightweight exoskeleton prototype focused on comfort and safety.
4. Test how well real-time muscle signals can drive movement.
5. Evaluate improvements in walking quality and user effort.

---

---

What You Will Do Step by Step

---

1) Learn key concepts about gait and EMG signals.
2) Build or simulate a basic exoskeleton that can apply assistive forces.
3) Collect EMG data from volunteers during walking tasks.
4) Create a simple real-time controller that maps EMG to assistance.
5) Run trials to observe safety, comfort, and performance outcomes.
6) Analyze metrics like speed, stability, and effort.
7) Discuss limitations and potential improvements.

---

---

Expected Outcome

Demonstration of a functional, user-responsive exoskeleton concept with data showing improved gait metrics and user experience, along with a clear plan for further refinement and real-world testing.