Smart exoskeleton-assisted gait rehabilitation for post-stroke patients using multimodal sensor fusion

Project Abstract

This study presents a smart exoskeleton-assisted gait rehabilitation system for post-stroke patients utilizing multimodal sensor fusion to deliver personalized, adaptive therapy with real-time feedback. The proposed platform integrates limb- and body-worne sensors, including inertial measurement units (IMUs), pressure sensors embedded in footwear, myoelectric signals from affected muscles, and holographic gait phase estimation via computer vision anchors, to capture comprehensive kinematic, kinetic, and neuromuscular data. A novel fusion framework combines model-based kinematics with data-driven neural networks to infer joint trajectories, gait symmetry, and propulsion metrics in real time, enabling precise assistance modulation. The exoskeleton employs actuated ankle, knee, and hip joints with impedance-based control and torque assist optimization guided by a patient-specific rehabilitation protocol that adapts to fatigue, spatiotemporal gait deviations, and motor learning progress. A dual-layer control loop ensures safety and responsiveness a fast inner loop governs joint-level impedance and trajectory tracking, while a slower outer loop updates assistance profiles and progression criteria using reinforcement learning-informed policies. The system supports task-oriented training paradigms, including plantigrade and heel-to-toe walking, obstacle negotiation, and treadmill/over-ground integration, to foster task specificity and transfer to daily activities. Data-driven gait event detection andPhase-of-Gait estimation enable seamless synchronization between user intention and exoskeleton actuation, reducing cognitive load and improving engagement. To evaluate efficacy, a randomized controlled trial with fifty post-stroke participants across subacute and chronic stages will compare the multimodal fusion-assisted exoskeleton against conventional physical therapy and single-sensor assisted therapy. Primary outcomes include improvements in comfortable gait speed, gait endurance (4-minute walk test), and temporal-spatial symmetry; secondary outcomes comprise Fugl-Meyer Assessment for lower extremity, Functional Ambulation Category, and user engagement metrics. The abstract system’s adaptability is demonstrated through a case series of differential motor impairments, illustrating how vessel-like neural coupling and proprioceptive feedback from the exoskeleton promote neuroplastic changes and motor relearning. A mixed-methods analysis will incorporate qualitative user experience, perceived exertion, and safety event reporting to inform iterative design refinements. Anticipated contributions include a scalable multimodal sensing architecture, a robust real-time fusion algorithm with generalizable gait models, and an adaptable rehabilitation framework that personalizes assistance while maximizing neurorehabilitation potential. The integration of multimodal data streams enables precise, real-time torque modulation and responsive therapeutic progression, aiming to accelerate functional recovery, enhance independence in activities of daily living, and reduce long-term caregiver burden. Potential limitations addressed include sensor drift, inter-user variability in neuromuscular patterns, and computational demands, with proposed mitigation through calibration routines, transfer learning, and edge-computing optimizations. Overall, the study advances a clinically viable, patient-centric approach to post-stroke gait rehabilitation, leveraging intelligent exoskeleton control and multimodal sensor fusion to optimize outcomes and engagement.

Project Overview

What This Project Is About

---

A plain-language overview of using a wearable robotic device (an exoskeleton) to help people recover walking after a stroke. The project studies how sensors collect body movement and muscle signals, and how a computer program combines these inputs to control the exoskeleton to assist walking in a safe, natural way. It also explores how to make the device comfortable, affordable, and easy to use in rehabilitation settings. The goal is to improve gait, balance, and confidence during recovery.

---

---

The Problem It Addresses

---

Many stroke survivors have trouble walking, which limits independence and increases health risks. Current rehab tools can be expensive, bulky, or require a therapist to guide every move. This project examines how a lighter, smarter exoskeleton supported by multiple sensors can provide personalized assistance, reduce therapist workload, and enable more effective, home-friendly rehabilitation.

---

---

Objectives of the Project

---

1. Review existing gait rehabilitation methods and sensor technologies.
2. Design a multimodal sensor system to capture gait data and muscle activity.
3. Develop control algorithms to synchronize exoskeleton assistance with user intent.
4. Test safety, comfort, and usability with mock trials and healthy volunteers.
5. Evaluate improvements in walking speed, symmetry, and balance in simulations.

---

---

What You Will Do Step by Step

---

1. Study literature on stroke rehab and exoskeletons to define requirements.
2. Prototype a sensor setup (e.g., motion, force, and muscle signals).
3. Develop software to process data and issue exoskeleton commands.
4. Conduct small-scale tests to refine safety features and comfort.
5. Run data analyses to measure gait improvements and system reliability.

---

---

Expected Outcome

A working concept of a multimodal sensor-guided exoskeleton for post-stroke gait rehab, with preliminary data showing improved walking metrics and practical recommendations for clinical use and future development.