In this paper, we present a generalized modeling tool for predicting the output force profile of vacuum-powered soft actuators using a simplified geometrical approach and the principle of virtual work. Previous work has derived analytical formulas to model the force-contraction profile of specific actuators. To enhance the versatility and the efficiency of the modelling process we propose a generalized numerical algorithm based purely on geometrical inputs, which can be tailored to the desired actuator, to estimate its force-contraction profile quickly and for any combination of varying geometrical parameters. We identify a class of linearly contracting vacuum actuators that consists of a polymeric skin guided by a rigid skeleton and apply our model to two such actuators-vacuum bellows and Fluid-driven Origami-inspired Artificial Muscles-to demonstrate the versatility of our model. We perform experiments to validate that our model can predict the force profile of the actuators using its geometric principles, modularly combined with design-specific external adjustment factors. Our framework can be used as a versatile design tool that allows users to perform parametric studies and rapidly and efficiently tune actuator dimensions to produce a force-contraction profile to meet their needs, and as a pre-screening tool to obviate the need for multiple rounds of time-intensive actuator fabrication and testing.
The majority of soft pneumatic actuators for rehabilitation exercises have been designed for adult users. Specifically, there is a paucity of soft rehabilitative devices designed for infants with upper and lower limb motor disabilities. We present a low-profile vacuum-powered artificial muscle (LP-VPAM) with dimensions suitable for infants. The actuator produced a maximum force of 26 N at vacuum pressures of −40 kPa. When implemented in an experimental model of an infant leg in an antagonistic-agonist configuration to measure resultant knee flexion, the actuator generated knee flexion angles of 43° and 61° in the prone and side-lying position, respectively.
This study explores two methods of predicting non-cyclic upper-body motions using classification algorithms. Exoskeletons currently face challenges with low fluency, hypothesized to be in part caused by the lag in active control innate in many leader–follower paradigms seen in today’s systems, leading to energetic inefficiencies and discomfort. To address this, we employ k-nearest neighbor (KNN) and deep learning models to predict motion characteristics, such as magnitude and category, from surface electromyography (sEMG) signals. Data were collected from six muscles located around the elbow. The sEMG signals were processed to identify significant activation changes. Two classification approaches were utilized: a KNN algorithm that categorizes motion based on the slopes of processed sEMG signals at change points and a deep neural network employing continuous categorization. Both methods demonstrated the capability to predict future voluntary non-cyclic motions up to and beyond commonly acknowledged electromechanical delay times, with the deep learning model able to predict, with certainty at or beyond 90%, motion characteristics even prior to myoelectric activation of the muscles involved. Our findings indicate that these classification algorithms can be used to predict upper-body non-cyclic motions to potentially increase machine interfacing fluency. Further exploration into regression-based prediction models could enhance the precision of these predictions, and further work could explore their effects on fluency when utilized in a tandem or wearable robotic application.
The use of surface electromyography (sEMG) for conventional motion classification and prediction has had limitations due to sensor hardware differences. With the popularization of deep learning-based approaches to the application of motion prediction, this study explores the effects that different hardware sensor platforms have on the performance of a deep learning neural network trained to predict the one-degree-of-freedom (DoF) angular trajectory of a human. Two different sEMG sensor platforms were used to collect raw data from subjects conducting exercises, which was used to train a neural network designed to predict the future angular trajectory of the arm. The results show that the raw data originating from different sensor hardware with different configurations (including the communication method, data acquisition unit (DAQ) usage, electrode configuration, buffering method, preprocessing method, and experimental variables like the sampling frequency) produced bi-LSTM networks that performed similarly. This points to the hardware-agnostic nature of such deep learning networks.
Today’s exoskeletons face challenges with low fluency (a quantifiable alternative to “seamlessness”), hypothesized to be caused by a lag in active control innate in many leader–follower paradigms seen in contemporary systems, leading to inefficiencies and discomfort. Furthermore, tandem fluency, a variation of fluency specific for tandem robots systems as exoskeletons, is yet to be rigorously tested in practice. This study aims to utilize metrics of tandem fluency in order to demonstrate improved human–robot interaction (HRI) in exoskeletons through human subject testing of a prototype 1 degree of freedom (DoF) exoskeleton using a motion prediction bidirectional long short-term memory (bi-LSTM) deep learning network. Subjects were recruited to conduct various upper body exercises about the elbow joint, and the collected sEMG, goniometer, and gas exchange data was used to design, test, optimize, and assess the performance of the 1 DoF exoskeleton using tandem fluency metrics. We found that the correlation between I-ACT, a metric of tandem fluency, the subjective survey responses, and metabolic data suggest that the use of a predictive bi-LSTM network to control a 1 DoF exoskeleton about the elbow results in an overall positive trend, which may correlate to high tandem fluency.