Augmenting Postural Stability in Older Adults Using a Lower Extremity Exoskeleton Controlled by Wearable Accelerometer and Surface Electromyographic Data

Falls are the second leading cause of unintentional injury-related deaths worldwide, with adults over the age of 60 accounting for the highest number of fatal incidents. This increased risk is largely due to age-related declines in muscle strength, neurological function, and overall mobility. This project is based on the concept of employing a six-degree-of-freedom lower-limb exoskeleton controlled via EMG signals to support balance recovery in the event of a perturbation during standing. In a model where the perturbation is simulated by a backward pull, the subject typically responds with a dorsiflexion movement in an attempt to counteract the disturbance. However, a physiological delay exists between the onset of the perturbation and the initiation of reactive muscular activity. Since the EMG signal can only be detected after this delay, the control strategy inherently includes a reaction time lag. This delay is then compounded by the time required for the exoskeleton system to process the EMG signal and convert it into an effective torque at the ankle joints. To address the limitation posed by physiological delay, we explore an alternative approach based on monitoring the subject’s inertial response during perturbation, an event that occurs before the onset of voluntary muscular activity. In this context, an acceleration-based controller offers a promising solution by enabling an earlier and potentially more effective intervention. This thesis focused on enhancing the existing EMG-based control system of a lower-limb exoskeleton and developing an alternative control strategy based on inertial data. The aim was to improve the responsiveness of the system in applying corrective ankle torques, thereby supporting balance recovery in the presence of external perturbations. The work addressed three critical aspects of the system’s performance: signal noise reduction, improvement of control accuracy, and minimization of latency. These enhancements were essential to achieving a real-time, reliable assistive response. The refinements made to the EMG-based controller led to notable improvements in signal quality, control performance, and latency reduction. The integration of a Kalman filter into the EMG-processing pipeline showed good predictive performance, although it did not significantly reduce processing delays. Nonetheless, the approach remains promising and deserves further optimization for real-time applications. As for the acceleration-based controller, the feasibility of estimating the ankle angle signal, derived from the inverted pendulum model with inertial signals as inputs, for driving the control action of the exoskeleton was demonstrated. However, the absence of gyroscopic data limited its ability to capture the system’s dynamic behavior. Future work will focus on integrating an inertial measurement unit (IMU) and developing a dedicated processing pipeline. The exoskeleton will need to be tested on a subject to validate our findings in terms of performance. Ultimately, the next step will be a comprehensive comparison between the EMG-based and IMU-based strategies, evaluating both technical performance and user perception during balance recovery. This work lays the groundwork for more adaptive and intuitive control solutions in wearable robotics for fall prevention.