This work studies upper-limb impairment resulting from stroke or traumatic brain injury and presents a simple technological solution for a subset of patients: a soft, active stretching aid for at-home use. To better understand the issues associated with existing associated rehabilitation devices, customer discovery conversations were conducted with 153 people in the healthcare ecosystem (60 patients, 30 caregivers, and 63 medical providers). These patients fell into two populations: spastic (stiff, clenched hands) and flaccid (limp hands). Focusing on the first category, a set of design constraints was developed based on the information collected from the customer discovery. With these constraints in mind, a powered wrist-hand stretching orthosis (exoskeleton) was designed and prototyped as a preclinical study (T0 basic science research) to aid in recovery. The orthosis was tested on two patients for proof-of-concept, one survivor of stroke and one of traumatic brain injury. The prototype was able to consistently open both patients’ hands. A mathematical model was developed to characterize joint stiffness based on experimental testing. Donning and doffing times for the prototype averaged 76 and 12.5 s, respectively, for each subject unassisted. This compared favorably to times shown in the literature. This device benefits from simple construction and low-cost materials and is envisioned to become a therapy device accessible to patients in the home. This work lays the foundation for phase 1 clinical trials and further device development.

Orthotic wrist supports will be beneficial for people with muscular weakness to keep their hand in a neutral rest position and prevent potential wrist contractures. Compensating the weight of the hands is complex since the level of support depends on both wrist and forearm orientations. To explore simplified approaches, two different weight compensation strategies (constant and linear) were compared to the theoretical ideal sinusoidal profile and no compensation in eight healthy subjects using a mechanical wrist support system. All three compensation strategies showed a significant reduction of 47–53% surface electromyography activity in the anti-gravity m. extensor carpi radialis. However, for the higher palmar flexion region, a significant increase of 44–61% in the m. flexor carpi radialis was found for all compensation strategies. No significant differences were observed between the various compensation strategies. Two conclusions can be drawn: (1) a simplified torque profile (e.g., constant or linear) for weight compensation can be considered as equally effective as the theoretically ideal sinusoidal profile and (2) even the theoretically ideal profile provides no perfect support as other factors than weight, such as passive joint impedance, most likely influence the required compensation torque for the wrist joint.

Assistive forces transmitted from wearable robots to the robot’s users are often defined by controllers that rely on the accurate estimation of the human posture. The compliant nature of the human–robot interface can negatively affect the robot’s ability to estimate the posture. In this article, we present a novel algorithm that uses machine learning to correct these errors in posture estimation. For that, we recorded motion capture data and robot performance data from a group of participants (n = 8; 4 females) who walked on a treadmill while wearing a wearable robot, the Myosuit. Participants walked on level ground at various gait speeds and levels of support from the Myosuit. We used optical motion capture data to measure the relative displacement between the person and the Myosuit. We then combined this data with data derived from the robot to train a model, using a grading boosting algorithm (XGBoost), that corrected for the mechanical compliance errors in posture estimation. For the Myosuit controller, we were particularly interested in the angle of the thigh segment. Using our algorithm, the estimated thigh segment’s angle RMS error was reduced from 6.3° (2.3°) to 2.5° (1.0°), mean (standard deviation). The average maximum error was reduced from 13.1° (4.9°) to 5.9° (2.1°). These improvements in posture estimation were observed for all of the considered assistance force levels and walking speeds. This suggests that ML-based algorithms provide a promising opportunity to be used in combination with wearable-robot sensors for an accurate user posture estimation.