Walking Robot Control with a Machine Learning-based Ground Reaction Force Predictor and Generated Linear Contact Model*

2021 ◽  
Author(s):  
Sergei Savin ◽  
Svyatoslav Golousov ◽  
Eduard Zalyaev ◽  
Alek Salikhzyanov ◽  
Alexandr Klimchik
Keyword(s):  
Machine Learning ◽  
Ground Reaction Force ◽  
Robot Control ◽  
Reaction Force ◽  
Contact Model ◽  
Walking Robot

scholarly journals A Smart Terrain Identification Technique Based on Electromyography, Ground Reaction Force, and Machine Learning for Lower Limb Rehabilitation

2020 ◽  
Vol 10 (8) ◽  
pp. 2638 ◽  
Author(s):  
Shuo Gao ◽  
Yixuan Wang ◽  
Chaoming Fang ◽  
Lijun Xu
Keyword(s):  
Machine Learning ◽  
Ground Reaction Force ◽  
Lower Limb ◽  
Confusion Matrix ◽  
Reaction Force ◽  
Simple System ◽  
Support Vector ◽  
Terrain Classification ◽  
Limb Rehabilitation ◽  
Lower Limb Rehabilitation

Automatic terrain classification in lower limb rehabilitation systems has gained worldwide attention. In this field, a simple system architecture and high classification accuracy are two desired attributes. In this article, a smart neuromuscular–mechanical fusion and machine learning-based terrain classification technique utilizing only two electromyography (EMG) sensors and two ground reaction force (GRF) sensors is reported for classifying three different terrains (downhill, level, and uphill). The EMG and GRF signals from ten healthy subjects were collected, preprocessed and segmented to obtain the EMG and GRF profiles in each stride, based on which twenty-one statistical features, including 9 GRF features and 12 EMG features, were extracted. A support vector machine (SVM) machine learning model is established and trained by the extracted EMG features, GRF features and the fusion of them, respectively. Several methods or statistical metrics were used to evaluate the goodness of the proposed technique, including a paired-t-test and Kruskal–Wallis test for correlation analysis of the selected features and ten-fold cross-validation accuracy, confusion matrix, sensitivity and specificity for the performance of the SVM model. The results show that the extracted features are highly correlated with the terrain changes and the fusion of the EMG and GRF features produces the highest accuracy of 96.8%. The presented technique allows simple system construction to achieve the precise detection of outcomes, potentially advancing the development of terrain classification techniques for rehabilitation.

Dynamically adjustable foot-ground contact model to estimate ground reaction force during walking and running

2016 ◽  
Vol 45 ◽  
pp. 62-68 ◽  
Author(s):  
Yihwan Jung ◽  
Moonki Jung ◽  
Jiseon Ryu ◽  
Sukhoon Yoon ◽  
Sang-Kyoon Park ◽  
...  
Keyword(s):  
Ground Reaction Force ◽  
Reaction Force ◽  
Contact Model ◽  
Ground Contact

scholarly journals Indirect Estimation of Vertical Ground Reaction Force from a Body-Mounted INS/GPS Using Machine Learning

Sensors ◽  
2021 ◽  
Vol 21 (4) ◽  
pp. 1553
Author(s):  
Dharmendra Sharma ◽  
Pavel Davidson ◽  
Philipp Müller ◽  
Robert Piché
Keyword(s):  
Machine Learning ◽  
Ground Reaction Force ◽  
Short Term Memory ◽  
Reaction Force ◽  
Training Data ◽  
Machine Learning Techniques ◽  
Vertical Ground Reaction Force ◽  
Indirect Estimation ◽  
Gait Parameters ◽  
Vertical Ground

Vertical ground reaction force (vGRF) can be measured by force plates or instrumented treadmills, but their application is limited to indoor environments. Insoles remove this restriction but suffer from low durability (several hundred hours). Therefore, interest in the indirect estimation of vGRF using inertial measurement units and machine learning techniques has increased. This paper presents a methodology for indirectly estimating vGRF and other features used in gait analysis from measurements of a wearable GPS-aided inertial navigation system (INS/GPS) device. A set of 27 features was extracted from the INS/GPS data. Feature analysis showed that six of these features suffice to provide precise estimates of 11 different gait parameters. Bagged ensembles of regression trees were then trained and used for predicting gait parameters for a dataset from the test subject from whom the training data were collected and for a dataset from a subject for whom no training data were available. The prediction accuracies for the latter were significantly worse than for the first subject but still sufficiently good. K-nearest neighbor (KNN) and long short-term memory (LSTM) neural networks were then used for predicting vGRF and ground contact times. The KNN yielded a lower normalized root mean square error than the neural network for vGRF predictions but cannot detect new patterns in force curves.

scholarly journals An Exploration of Machine-Learning Estimation of Ground Reaction Force from Wearable Sensor Data

Sensors ◽  
2020 ◽  
Vol 20 (3) ◽  
pp. 740
Author(s):  
Danica Hendry ◽  
Ryan Leadbetter ◽  
Kristoffer McKee ◽  
Luke Hopper ◽  
Catherine Wild ◽  
...  
Keyword(s):  
Machine Learning ◽  
Ground Reaction Force ◽  
Reaction Force ◽  
Sensor Data ◽  
Wearable Sensor ◽  
Learning Models ◽  
Data Set ◽  
Sensor Model ◽  
Single Sensor ◽  
Machine Learning Models

This study aimed to develop a wearable sensor system, using machine-learning models, capable of accurately estimating peak ground reaction force (GRF) during ballet jumps in the field. Female dancers (n = 30) performed a series of bilateral and unilateral ballet jumps. Dancers wore six ActiGraph Link wearable sensors (100 Hz). Data were collected simultaneously from two AMTI force platforms and synchronised with the ActiGraph data. Due to sensor hardware malfunctions and synchronisation issues, a multistage approach to model development, using a reduced data set, was taken. Using data from the 14 dancers with complete multi-sensor synchronised data, the best single sensor was determined. Subsequently, the best single sensor model was refined and validated using all available data for that sensor (23 dancers). Root mean square error (RMSE) in body weight (BW) and correlation coefficients (r) were used to assess the GRF profile, and Bland–Altman plots were used to assess model peak GRF accuracy. The model based on sacrum data was the most accurate single sensor model (unilateral landings: RMSE = 0.24 BW, r = 0.95; bilateral landings: RMSE = 0.21 BW, r = 0.98) with the refined model still showing good accuracy (unilateral: RMSE = 0.42 BW, r = 0.80; bilateral: RMSE = 0.39 BW, r = 0.92). Machine-learning models applied to wearable sensor data can provide a field-based system for GRF estimation during ballet jumps.

scholarly journals Running ground reaction forces across footwear conditions are predicted from the motion of two body mass components

2019 ◽  
Vol 126 (5) ◽  
pp. 1315-1325 ◽  
Author(s):  
Andrew B. Udofa ◽  
Kenneth P. Clark ◽  
Laurence J. Ryan ◽  
Peter G. Weyand
Keyword(s):  
Ground Reaction Force ◽  
Lower Limb ◽  
Reaction Force ◽  
Ground Reaction Forces ◽  
Vertical Ground Reaction Force ◽  
Time Patterns ◽  
Foot Strike ◽  
Reaction Forces ◽  
Vertical Ground ◽  
Force Time

Although running shoes alter foot-ground reaction forces, particularly during impact, how they do so is incompletely understood. Here, we hypothesized that footwear effects on running ground reaction force-time patterns can be accurately predicted from the motion of two components of the body’s mass (mb): the contacting lower-limb (m1 = 0.08mb) and the remainder (m2 = 0.92mb). Simultaneous motion and vertical ground reaction force-time data were acquired at 1,000 Hz from eight uninstructed subjects running on a force-instrumented treadmill at 4.0 and 7.0 m/s under four footwear conditions: barefoot, minimal sole, thin sole, and thick sole. Vertical ground reaction force-time patterns were generated from the two-mass model using body mass and footfall-specific measures of contact time, aerial time, and lower-limb impact deceleration. Model force-time patterns generated using the empirical inputs acquired for each footfall matched the measured patterns closely across the four footwear conditions at both protocol speeds ( r2 = 0.96 ± 0.004; root mean squared error  = 0.17 ± 0.01 body-weight units; n = 275 total footfalls). Foot landing angles (θF) were inversely related to footwear thickness; more positive or plantar-flexed landing angles coincided with longer-impact durations and force-time patterns lacking distinct rising-edge force peaks. Our results support three conclusions: 1) running ground reaction force-time patterns across footwear conditions can be accurately predicted using our two-mass, two-impulse model, 2) impact forces, regardless of foot strike mechanics, can be accurately quantified from lower-limb motion and a fixed anatomical mass (0.08mb), and 3) runners maintain similar loading rates (ΔFvertical/Δtime) across footwear conditions by altering foot strike angle to regulate the duration of impact. NEW & NOTEWORTHY Here, we validate a two-mass, two-impulse model of running vertical ground reaction forces across four footwear thickness conditions (barefoot, minimal, thin, thick). Our model allows the impact portion of the impulse to be extracted from measured total ground reaction force-time patterns using motion data from the ankle. The gait adjustments observed across footwear conditions revealed that runners maintained similar loading rates across footwear conditions by altering foot strike angles to regulate the duration of impact.

scholarly journals Small changes in ball position at address cause a chain effect in golf swing

2021 ◽  
Vol 11 (1) ◽  
Author(s):  
Sung Eun Kim ◽  
Jangyun Lee ◽  
Sae Yong Lee ◽  
Hae-Dong Lee ◽  
Jae Kun Shim ◽  
...  
Keyword(s):  
Ground Reaction Force ◽  
Reaction Force ◽  
Statistical Parametric Mapping ◽  
Golf Swing ◽  
Whole Body ◽  
Centre Of Mass ◽  
Body Segment ◽  
A Chain ◽  
Professional Golfers ◽  
The Right

AbstractThe purpose of this study was to investigate how the ball position along the mediolateral (M-L) direction of a golfer causes a chain effect in the ground reaction force, body segment and joint angles, and whole-body centre of mass during the golf swing. Twenty professional golfers were asked to complete five straight shots for each 5 different ball positions along M-L: 4.27 cm (ball diameter), 2.14 cm (ball radius), 0 cm (reference position at preferred ball position), – 2.14 cm, and – 4.27 cm, while their ground reaction force and body segment motions were captured. The dependant variables were calculated at 14 swing events from address to impact, and the differences between the ball positions were evaluated using Statistical Parametric Mapping. The left-sided ball positions at address showed a greater weight distribution on the left foot with a more open shoulder angle compared to the reference ball position, whereas the trend was reversed for the right-sided ball positions. These trends disappeared during the backswing and reappeared during the downswing. The whole-body centre of mass was also located towards the target for the left-sided ball positions throughout the golf swing compared to the reference ball position, whereas the trend was reversed for the right-sided ball positions. We have concluded that initial ball position at address can cause a series of chain effects throughout the golf swing.

Sign in / Sign up

Export Citation Format

Share Document