IEEE ... International Conference on Rehabilitation Robotics : [proceedings]最新文献

Predicting grip force and wrist angle during dynamic hand movements is crucial for advancing upper-limb prosthetic systems, enabling simultaneous and proportional control of multiple degrees of freedom (DOFs). This study introduces a novel wearable ultrasound-based system that leverages M-mode data from four single-element transducers placed on the forearm to capture muscle activity for the concurrent prediction of grip force and wrist angle. A multi-layer perceptron (MLP) regressor was utilized for the simultaneous prediction of both parameters, and a comparative analysis was conducted using a Gaussian process regressor (GPR), which is commonly adopted previously in similar studies. The system was validated on unseen data from five participants without limb loss. The MLP demonstrated superior performance compared to GPR, achieving $mathbf{R}^{mathbf{2}}$ values of $0.85 pm 0.06$ for wrist angle prediction and $0.74 pm 0.07$ for grip force. These findings underscore the challenges of predicting simultaneous grip force and wrist angle during dynamic hand movements and highlight the need to address these issues for intuitive and practical prosthetic control in real-world scenarios.

State-of-the-art upper-limb myoelectric prostheses are typically controlled using classification-based models that do not offer simultaneous control of wrist and hand movements (degrees of freedom or DOFs). Regression-based alternatives are being studied because they do offer simultaneous DOF control, yielding more natural movements, but generally require longer training routines. We investigated methods to reduce the training burden for regression-based myoelectric control. Five different methods to train regression models were tested using electromyographic (EMG) data collected from the forearms of 10 able-bodied participants. First, models were either trained traditionally with data from elbows at 90° position, with data from 3 limb positions, or trained by few-shot learning (with fewer data from 3 limb positions). Then, transfer learning was employed to pre-train models using data from all other users, with the models subsequently fine-tuned using either traditional or few-shot learning with new end-user data. The resulting five models were evaluated using linear regressor-, Convolutional Neural Network-, and Transformer-based approaches. Interestingly, the transfer learning pre-trained model in conjunction with few-shot fine-tuning achieved the second-highest median $mathbf{R}^{2}$ of 0.76 across all participants. Our findings offer a proof of concept for regression-based myoelectric control of multiple DOFs that may more closely resemble natural limb function.

Hand function assessment enables the therapist to select the proper rehabilitation therapy as well as to monitor the recovery progress. The Virtual Eggs Test (VET) stands out from the other tests for hand dexterity evaluation since evaluates both gross and fine dexterity. This is achieved by integrating measures of accuracy and speed in transporting objects with increasing fragilities. A clinical trial to validate VET was carried out with a population of amputees and a small group of healthy participants, showing promising results. To improve the resolution of the test, here we propose a revised version of the execution protocol without extending the duration. The new version has been preliminary evaluated by seven healthy participants of different ages. The results of this pilot study support test-retest reliability for the original formulation of the Fine Dexterity Index, while suggesting for a revision of the Gross Dexterity Index.

Stroke causes severe tactile deficiencies which affects motor control when grasping and lifting objects. Understanding the fundamental neural disorders associated with tactile deficits is crucial to developing rehabilitation and treatment plans correspondingly. Earlier studies have studied the dynamics between finger grasping behavior and arm muscle activation in stroke patients. However, the exact neuromuscular synergy of tactile perception and arm usage is left unexplored. Here we designed a comprehensive experiment platform and tested the potential synergy on 12 healthy young adults, serving as a control group to establish a foundation for future studies on stroke patients. The experimental platform consists of a lever arm on which torques can be applied to the subject's arm. The end effector is equipped with a special ultrasonic friction modulation plate that can reduce the apparent friction of the object by up to 63 %, simulating real-world grasping tasks in a controlled setting. The experiments were performed under varying conditions of friction and arm usage. Results indicate significant effects of tactile stimulation on grasping force adaptation ($p<0.05$ in 8 of 12 experimental conditions). In the meantime, arm usage did not show a significant synergy with tactile perception ($p=0.44$ in grasping force adaptation amplitude, and $p=0.73$ in reflex delay). These findings demonstrate that the experimental platform can provide insights into human tactile behaviors, which is critical for studying tactile sensory and motor control synergy. The results will lay the groundwork for future research on underlying pathologies and rehabilitation strategies for stroke patients.

In the context of FES-based rehabilitation and assistance tasks, there is a growing interest in managing the induced fatigue, which is an obvious limiting factor in training duration. However, it is unclear whether and how fatigue may develop in tasks commonly used in stroke rehabilitation. This work explored the effect of FES- or volition- induced movements on EMG-based effort and M-wave metrics that characterise fatigue, during a continuous wrist target-tracking task completed by 22 unimpaired participants. We found no significant changes in the mechanical and electrical responses of the muscles during the Volition-only, FES-only, and VolitionFES conditions, suggesting that during this task, FES did not induce muscular fatigue, while both mental and physical demands were reported as low. Our results thus suggest that it may not be necessary to consider FES-induced fatigue during such continuous FES-assisted tasks, that can be used in poststroke motor rehabilitation training.

Rehabilitation therapy can be more effective and engaging when interactive technologies are involved. To enhance this experience, we integrated a haptic/biometric-based (HBB) Dynamic Difficulty Adjustment (DDA) system into the enAblegames ${}^{text {TM}}$ platform, which already uses body tracking for therapeutic gaming. This system adapts game and haptic difficulty in real time based on each patient's biometric data and performance, making therapy more personalized. We tested this system with 11 participants, comparing their experiences with and without DDA. The results were promising-36% preferred DDA-enhanced games, compared to just $mathbf{7 %}$ for non-DDA, and in a single-game scenario, preference for DDA increased by 50 %. These early findings suggest that HBB-DDA can make rehabilitation more engaging and tailored to individual needs. While more research is needed to understand its full impact, this system has the potential to improve patient experience and therapy outcomes.

For children with motor disabilities, a wide range of assistive technologies (such as exoskeletons, treadmill trainers, and body-weight support harnesses) exist to support learning to walk. However, after the onset of independent walking, few technologies are geared toward helping children with motor disabilities to practice walking and improve walking control. In this paper, we assess the ability of GoBot, a custom assistive robot with multiple game modes, to encourage one child with a motor disability to improve their amount, speed, and control of independent walking. We conducted a 12-session single-subject study and found that the child walked more and faster while engaging in lightly competitive races against a directly teleoperated GoBot, compared to during experiences in a standard of care condition. As a second and more exploratory element of our work, we equipped GoBot to autonomously play the common children's game red light, green light (RLGL) with the user as an entertaining way to motivate balance practice. Anecdotally, this RLGL activity led to some of the highest levels of child engagement. The preliminary findings of our single-subject study can benefit researchers working with assistive robots and physical therapists working with children with independent walking practice goals.

Dextrous robotic hand designs are pivotal in advancing the field of hand prosthetics, enabling users to perform robust grasping tasks essential for daily activities and improved quality of life. However, the widespread adoption of such dextrous hand prosthetics remains hindered by prohibitive costs that arise from the high number of actuators used to replicate natural human hand movements. This research seeks to address this challenge by critically evaluating the necessity of replicating many human hand DoFs and instead proposing a design that achieves 95% of human grasp configurations with a significantly reduced actuator count compared to conventional biomimetic robotic hands. We present the Hybrid-Actuated dextrous Anthropomorphic Robotic (HADAR) Hand, a novel 13-DoF robotic hand optimised for dextrous object grasping with a high degree of visual anthropomorphism. The HADAR Hand employs a hybrid actuation strategy, combining linkagedriven mechanisms for robust proximal joint control and tendon-based actuation for compliant distal joint movement, which is essential for delicate object handling. Utilizing insights from long-term occupational studies and Functional Range of Motion (FROM) data, rigid and elastomeric couplings were developed to bring down the actuator count to less than half the number of major tendons present in the human hand without compromising on the ability to perform human-like grasp configurations. Comprehensive performance evaluations validate the HADAR Hand's capabilities, with results demonstrating success in replicating hand grasps as per Cutkosky's taxonomy (14/15), Feix's GRASP taxonomy (31/33), and the Kapandji thumb opposability test (6/10). To promote accessibility and reproducibility, the HADAR Hand leverages cost-effective, widely available actuators, additive manufacturing techniques, and a compact, two-layer PCB capable of concurrently driving all 13 N20 DC motors that drive the HADAR Hand. This work represents a significant step in developing affordable, highperformance prosthetics and robotic hands.

Hybrid Brain-Computer Interfaces (hBCI) integrate brain and muscle signals to enhance motor rehabilitation of stroke survivors, by closing the loop between the lesioned brain and the paretic limb. To date, little attention has been devoted to their potential efficacy in managing the maladaptive movement patterns that afflict post-stroke motor outcome (unwanted abnormal co-contrations, spasticity). This study proposes a comparison of Cortico-Muscular Coherence (CMC) patterns assessed in stroke patients before and after a 1-month rehabilitation intervention based on a hBCI-controlled Functional Electrical Stimulation (FES) treatment, which included a module to monitor non-physiological movement patterns. Results demonstrated the efficacy of this type of assistive technology for post-stroke rehabilitation, addressing patient-tailored interventions able to reduce the maladaptive mechanisms.

Mass-produced industrial and service robots have been used to build therapeutic rehabilitation robots since the 1990s. REHAROB 1.0, the dual-arm upper limb rehabilitation robot prototype, was among the pioneers. The industrial collaborative robot is a new type of robot in terms of safety, appearance, performance, and cost, thanks to which it is proving itself in more and more applications. To take advantage of these benefits, the REHAROB 3.0 Robotic Trainer for Activities of Daily Living, also updated in its name, was designed and built using two industrial collaborative robots and two uniquely developed robotic fingers. The goal of our study was to introduce an iterative design and simulation process in which digital human modeling is intertwined with digital robot modeling, leading to an upper limb rehabilitation robot system targeting outstanding performance features such as synchronous exercising of every upper limb anatomical joint and training reaching and grasping Activities of Daily Living with real objects in a real environment.