Robust exosuit controller design for dynamic locomotor tasks

For healthy young adults, we investigated a way to reduce the energetic cost of both walking and running using a single robotic device. Initially, we conducted experiments to explore the metabolic impact of assistance profiles during running. A study with a tethered hip extension exosuit demonstrated that an assistance profile, inspired by a musculoskeletal simulation, can reduce the metabolic cost of running by 5.4% compared to running without wearing a suit, marking the first reduction in running at the time of its publication. Since this running profile differed from the walking assistance profile in the field, a subsequent study was conducted to develop an online algorithm for detecting walking and running gaits. This algorithm was then used to apply the appropriate assistance profile based on the wearer’s gait mode (i.e., walking vs. running) in a portable hip extension exosuit, and it was demonstrated for the first time that a single device could reduce the energetic cost of both walking (9.3%) and running (4.0%) when compared to locomotion without the exosuit.

Optimization and personalization of exoskeleton assistance: Comparing control architectures

Building upon the idea of a tailored assistance profile for each gait mode, our research explored the impact of personalized assistance for individual users to address inter-subject variability in response to a fixed assistance profile. The underlying hypothesis was that an assistive profile that benefits one participant may hinder another. To test this hypothesis, we leveraged a human-in-the-loop optimization technique, where metabolic cost is estimated in real-time, and controller parameters are updated until metabolic cost is minimized. In one study with a tethered multi-articular exosuit, we applied a Bayesian optimization algorithm to personalize assistance for ankle plantarflexion and hip extension. The findings demonstrated that optimized assistance could reduce metabolic costs by 36% compared to walking without the suit. In another study with a tethered hip flexion exosuit, we used the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) algorithm and demonstrated that individualized assistance tailored to each participant significantly reduces the energetic cost of walking, surpassing the benefits of a fixed, generic assistance profile. Furthermore, beyond addressing inter-subject variability, optimizing exoskeleton assistance can enable better comparisons across devices or control approaches because the best versions in each scenario can be found and compared. In a study with a frontal-plane hip exoskeleton, we optimized assistance in both torque and position control schemes, demonstrating that neither of these controllers appeared promising in reducing the metabolic cost of walking. These results highlight the significance of personalization and optimization of exoskeleton assistance.

Adapting wearable robot technology for clinical populations

Leveraging the knowledge gleaned from experimental work with healthy individuals, we explored the potential of wearable robot technology in addressing gait impairments among individuals with Parkinson’s disease (PD). By conducting a series of single-subject experiments, a proof-of-concept study demonstrated, for the first time, that hip flexion assistance can prevent freezing of gait in PD. Additionally, a separate study involving two PD participants showed that hip flexion assistance could also improve short, shuffling steps, although it did not significantly impact walking speed. These foundational studies, while preliminary, underscore the potential of exosuit technology in alleviating the most challenging and distressing motor features in Parkinson’s disease.