Impact of Haptic Cues and an Active Ankle Exoskeleton on Gait Characteristics.

OBJECTIVE: This study examined the interaction of gait-synchronized vibrotactile cues with an active ankle exoskeleton that provides plantarflexion assistance. BACKGROUND: An exoskeleton that augments gait may support collaboration through feedback to the user about the state of the exoskeleton or characteristics of the task. METHODS: Participants (N = 16) were provided combinations of torque assistance and vibrotactile cues at pre-specified time points in late swing and early stance while walking on a self-paced treadmill. Participants were either given explicit instructions (N = 8) or were allowed to freely interpret (N=8) how to coordinate with cues. RESULTS: For the free interpretation group, the data support an 8% increase in stride length and 14% increase in speed with exoskeleton torque across cue timing, as well as a 5% increase in stride length and 7% increase in speed with only vibrotactile cues. When given explicit instructions, participants modulated speed according to cue timing-increasing speed by 17% at cues in late swing and decreasing speed 11% at cues in early stance compared to no cue when exoskeleton torque was off. When torque was on, participants with explicit instructions had reduced changes in speed. CONCLUSION: These findings support that the presence of torque mitigates how cues were used and highlights the importance of explicit instructions for haptic cuing. Interpreting cues while walking with an exoskeleton may increase cognitive load, influencing overall human-exoskeleton performance for novice users. APPLICATION: Interactions between haptic feedback and exoskeleton use during gait can inform future feedback designs to support coordination between users and exoskeletons.

Assessment of a powered ankle exoskeleton on human stability and balance.

Wearable robotic systems, such as exoskeletons, are designed to assist human motion; however, they are typically only studied during level walking. Before exoskeletons are broadly integrated into unstructured environments, it will be important to evaluate exoskeletons in a broader set of relevant tasks. A balance beam traverse was used to represent a constrained foot placement task for examining balance and stability. Participants (n = 17) completed the task in their own shoes (Pre-Exoskeleton and Post-Exoskeleton trials), and when wearing a lower-limb exoskeleton (Dephy ExoBoot) in both powered and unpowered states. Data were collected via inertial measurement units (on the torso and feet) and analyzed on a pooled level (with data from all participants) and on an individual level (participant-specific confidence intervals). When examining pooled data, it was observed that the exoskeleton had mixed effects on stride stability metrics. When compared to the Post-Exoskeleton shoe control, it was observed that stride duration was increased when wearing the exoskeleton (both powered and unpowered states), while normalized stride length and stride speed were not affected. Despite the changes in stride stability, overall balance (as measured by torso sway) remained unaffected by exoskeleton state. On an individual level, it was observed that not all participants followed these general trends, and within each metric, some increased, some decreased, and some had no change in the Powered Exoskeleton condition when compared to the Post-Exoskeleton Shoe condition: normalized stride length (0% increased, 12% decreased, 88% no change), stride duration (35% increased, 0% decreased, 65% no change), and torso sway (0% increased, 12% decreased, 88% no change). Our findings suggest that the lower-limb exoskeleton evaluated can be used during tasks that require balancing, and we recommend that balancing tasks be included in standards for exoskeleton evaluation.

Utility of Inter-subject Transfer Learning for Wearable-Sensor-Based Joint Torque Prediction Models.

Generalizability between individuals and groups is often a significant hurdle in model development for human subjects research. In the domain of wearable-sensor-controlled exoskeleton devices, the ability to generalize models across subjects or fine-tune more general models to individual subjects is key to enabling widespread adoption of these technologies. Transfer learning techniques applied to machine learning models afford the ability to apply and investigate the viability and utility such knowledge-transfer scenarios. This paper investigates the utility of single- and multi-subject based parameter transfer on LSTM models trained for "sensor-to-joint torque" prediction tasks, with regards to task performance and computational resources required for network training. We find that parameter transfer between both single- and multi-subject models provide useful knowledge transfer, with varying results across specific "source" and "target" subject pairings. This could be leveraged to lower model training time or computational cost in compute-constrained environments or, with further study to understand causal factors of the observed variance in performance across source and target pairings, to minimize data collection and model retraining requirements to select and personalize a generic model for personalized wearable-sensor-based joint torque prediction technologies.

A Neural Network Estimation of Ankle Torques From Electromyography and Accelerometry.

Estimations of human joint torques can provide clinically valuable information to inform patient care, plan therapy, and assess the design of wearable robotic devices. Predicting joint torques into the future can also be useful for anticipatory robot control design. In this work, we present a method of mapping joint torque estimates and sequences of torque predictions from motion capture and ground reaction forces to wearable sensor data using several modern types of neural networks. We use dense feedforward, convolutional, neural ordinary differential equation, and long short-term memory neural networks to learn the mapping for ankle plantarflexion and dorsiflexion torque during standing, walking, running, and sprinting, and consider both single-point torque estimation, as well as the prediction of a sequence of future torques. Our results show that long short-term memory neural networks, which consider incoming data sequentially, outperform dense feedforward, neural ordinary differential equation networks, and convolutional neural networks. Predictions of future ankle torques up to 0.4 s ahead also showed strong positive correlations with the actual torques. The proposed method relies on learning from a motion capture dataset, but once the model is built, the method uses wearable sensors that enable torque estimation without the motion capture data.

Ablation Analysis to Select Wearable Sensors for Classifying Standing, Walking, and Running.

The field of human activity recognition (HAR) often utilizes wearable sensors and machine learning techniques in order to identify the actions of the subject. This paper considers the activity recognition of walking and running while using a support vector machine (SVM) that was trained on principal components derived from wearable sensor data. An ablation analysis is performed in order to select the subset of sensors that yield the highest classification accuracy. The paper also compares principal components across trials to inform the similarity of the trials. Five subjects were instructed to perform standing, walking, running, and sprinting on a self-paced treadmill, and the data were recorded while using surface electromyography sensors (sEMGs), inertial measurement units (IMUs), and force plates. When all of the sensors were included, the SVM had over 90% classification accuracy using only the first three principal components of the data with the classes of stand, walk, and run/sprint (combined run and sprint class). It was found that sensors that were placed only on the lower leg produce higher accuracies than sensors placed on the upper leg. There was a small decrease in accuracy when the force plates are ablated, but the difference may not be operationally relevant. Using only accelerometers without sEMGs was shown to decrease the accuracy of the SVM.

Implementation of a Surface Electromyography-Based Upper Extremity Exoskeleton Controller Using Learning from Demonstration.

Upper-extremity exoskeletons have demonstrated potential as augmentative, assistive, and rehabilitative devices. Typical control of upper-extremity exoskeletons have relied on switches, force/torque sensors, and surface electromyography (sEMG), but these systems are usually reactionary, and/or rely on entirely hand-tuned parameters. sEMG-based systems may be able to provide anticipatory control, since they interface directly with muscle signals, but typically require expert placement of sensors on muscle bodies. We present an implementation of an adaptive sEMG-based exoskeleton controller that learns a mapping between muscle activation and the desired system state during interaction with a user, generating a personalized sEMG feature classifier to allow for anticipatory control. This system is robust to novice placement of sEMG sensors, as well as subdermal muscle shifts. We validate this method with 18 subjects using a thumb exoskeleton to complete a book-placement task. This learning-from-demonstration system for exoskeleton control allows for very short training times, as well as the potential for improvement in intent recognition over time, and adaptation to physiological changes in the user, such as those due to fatigue.

Classification of Anticipatory Signals for Grasp and Release from Surface Electromyography.

Surface electromyography (sEMG) is a technique for recording natural muscle activation signals, which can serve as control inputs for exoskeletons and prosthetic devices. Previous experiments have incorporated these signals using both classical and pattern-recognition control methods in order to actuate such devices. We used the results of an experiment incorporating grasp and release actions with object contact to develop an intent-recognition system based on Gaussian mixture models (GMM) and continuous-emission hidden Markov models (HMM) of sEMG data. We tested this system with data collected from 16 individuals using a forearm band with distributed sEMG sensors. The data contain trials with shifted band alignments to assess robustness to sensor placement. This study evaluated and found that pattern-recognition-based methods could classify transient anticipatory sEMG signals in the presence of shifted sensor placement and object contact. With the best-performing classifier, the effect of label lengths in the training data was also examined. A mean classification accuracy of 75.96% was achieved through a unigram HMM method with five mixture components. Classification accuracy on different sub-movements was found to be limited by the length of the shortest sub-movement, which means that shorter sub-movements within dynamic sequences require larger training sets to be classified correctly. This classification of user intent is a potential control mechanism for a dynamic grasping task involving user contact with external objects and noise. Further work is required to test its performance as part of an exoskeleton controller, which involves contact with actuated external surfaces.

Carbon, metals, and grain size correlate with bacterial community structure in sediments of a high arsenic aquifer.

Bacterial communities can exert significant influence on the biogeochemical cycling of arsenic (As). This has globally important implications since As in drinking water affects the health of over 100 million people worldwide, including in the Ganges-Brahmaputra Delta region of Bangladesh where geogenic arsenic in groundwater can reach concentrations of more than 10 times the World Health Organization's limit. Thus, the goal of this research was to investigate patterns in bacterial community composition across gradients in sediment texture and chemistry in an aquifer with elevated groundwater As concentrations in Araihazar, Bangladesh. We characterized the bacterial community by pyrosequencing 16S rRNA genes from aquifer sediment samples collected at three locations along a groundwater flow path at a range of depths between 1.5 and 15 m. We identified significant differences in bacterial community composition between locations in the aquifer. In addition, we found that bacterial community structure was significantly related to sediment grain size, and sediment carbon (C), manganese (Mn), and iron (Fe) concentrations. Deltaproteobacteria and Chloroflexi were found in higher proportions in silty sediments with higher concentrations of C, Fe, and Mn. By contrast, Alphaproteobacteria and Betaproteobacteria were in higher proportions in sandy sediments with lower concentrations of C and metals. Based on the phylogenetic affiliations of these taxa, these results may indicate a shift to more Fe-, Mn-, and humic substance-reducers in the high C and metal sediments. It is well-documented that C, Mn, and Fe may influence the mobility of groundwater arsenic, and it is intriguing that these constituents may also structure the bacterial community.