insideOut: A Bio-Inspired Machine Learning Approach to Estimating Posture in Robots Driven by Compliant Tendons

Estimates of limb posture are critical for controlling robotic systems. This is generally accomplished with angle sensors at individual joints that simplify control but can complicate mechanical design and robustness. Limb posture should be derivable from each joint's actuator shaft angle but this is problematic for compliant tendon-driven systems where (i) motors are not placed at the joints and (ii) nonlinear tendon stiffness decouples the relationship between motor and joint angles. Here we propose a novel machine learning algorithm to accurately estimate joint posture during dynamic tasks by limited training of an artificial neural network (ANN) receiving motor angles and tendon tensions, analogous to biological muscle and tendon mechanoreceptors. Simulating an inverted pendulum—antagonistically-driven by motors and nonlinearly-elastic tendons—we compare how accurately ANNs estimate joint angles when trained with different sets of non-collocated sensory information generated via random motor-babbling. Cross-validating with new movements, we find that ANNs trained with motor angles and tendon tension data predict joint angles more accurately than ANNs trained without tendon tension. Furthermore, these results are robust to changes in network/mechanical hyper-parameters. We conclude that regardless of the tendon properties, actuator behavior, or movement demands, tendon tension information invariably improves joint angle estimates from non-collocated sensory signals.

The Auger TLP: Calibration of Minimum Bottom Tendon Tension Based On Field Measurements

The Auger Tension Leg Platform (TLP), which was installed in 1994, is Shell’s first TLP in the Gulf of Mexico (GoM). The Auger TLP was designed during the time when the industry had not yet been able to fully investigate the global dynamic characteristics of TLPs, especially the high frequency dynamic responses of tendons, and the design tensions of the Auger tendons were not calibrated to scaled wave basin model tests like the later TLP projects since the Auger TLP. Based on the accumulated experience from more than two decades’ operation and a number of studies conducted on the Auger TLP global performance, it is revealed that the Auger tendon tension is conservative given the current operational limit; however, the extra conservatism has not been fully quantified due to the lack of model test data. With the recorded Auger global motions and tendon tensions from the on-board measurement system, the performance of the Auger TLP in extreme storms is becoming fully unveiled by calibrating the analytical predictions (both time-domain analysis and frequency-domain analysis) with the measurement data. Thus, the objectives of this paper are (i) to calibrate the TLP minimum tendon tension design recipe based on the high-fidelity field measurement data from Tropical Storm Cindy 2017 and Hurricane Laura 2020 using both time-domain and frequency-domain simulations, and (ii) to propose the new allowable vertical center of gravity (VCG) and the new tendon pretensions for the Auger TLP for the extreme storm conditions. It is concluded that the current allowable VCG can be increased or the current required tendon pretension can be decreased without compromising the safety to the platform during the extreme storm conditions.

Comprehending Optimality of Finger Flexor Tendon Pulley System using Computational Analysis

It is rare that existing prosthetic/orthotic designs are based on kinetostatics of a biological finger, especially its tendon- pulley system (TPS). Whether a biological TPS is optimal for use as a reference, say for design purposes, and if so in what sense, is also relatively unknown. We expect an optimal TPS to yield high range of flexion while operating with lower tendon tension, bowstringing, and pulley stresses. To gain insight into the TPS designs, we present a parametric study which is then used to determine optimal TPS configurations for the flexor mechanism. A compliant, flexure-based computational model is developed and simulated using the pseudo rigid body method, with various combinations of pulley/tendon attachment point locations, pulley heights, and widths. Results suggest that three distinct types of TPS configurations corresponding to single stiff pulley, or two stiff pulleys, or one stiff and one flexible-inextensible pulley per phalange can be optimal. For a TPS configuration similar to a biological one, the distal pulleys on the proximal and intermediate phalanges need to be like flexible-inextensible string loops that effectively model the behavior of joint and cruciate pulleys. We reckon that a biological flexor TPS may have evolved to maximize flexion range with minimum possible actuation tension, bowstringing and pulley stress. Our findings may be useful in not only developing efficient hand devices, but also in improving TPS reconstruction surgery procedures.

A 4-year study of hamstring injury outcomes in elite track and field using the British Athletics rehabilitation approach

ObjectivesThe British Athletics Muscle Injury Classification (BAMIC) correlates with return to play in muscle injury. The aim of this study was to examine hamstring injury diagnoses and outcomes within elite track and field athletes following implementation of the British Athletics hamstring rehabilitation approach.MethodsAll hamstring injuries sustained by elite track and field athletes on the British Athletics World Class Programme between December 2015 and November 2019 that underwent an MRI and had British Athletics medical team prescribed rehabilitation were included. Athlete demographics and specific injury details, including mechanism of injury, self-reported gait phase, MRI characteristics and time to return to full training (TRFT) were contemporaneously recorded.Results70 hamstring injuries in 46 athletes (24 women and 22 men, 24.6±3.7 years) were included. BAMIC grade and the intratendon c classification correlated with increased TRFT. Mean TRFT was 18.6 days for the entire cohort. Mean TRFT for intratendon classifications was 34±7 days (2c) and 48±17 days (3c). The overall reinjury rate was 2.9% and no reinjuries were sustained in the intratendon classifications. MRI variables of length and cross-sectional (CSA) area of muscle oedema, CSA of tendon injury and loss of tendon tension were associated with TRFT. Longitudinal length of tendon injury, in the intratendon classes, was not associated with TRFT.ConclusionThe application of BAMIC to inform hamstring rehabilitation in British Athletics results in low reinjury rates and favourable TRFT following hamstring injury. The key MRI variables associated with longer recovery are length and CSA of muscle oedema, CSA of tendon injury and loss of tendon tension.

Modeling and Design Optimization of Robotic Hoses for 3D Printing of Cement

Modeling tendon tensions for applications of tendon-actuated continuum robots under significant loading is necessary for sizing motors, tendons, and other components to ensure that the robot can safely support its mass during operation. While models exist that express tendon tensions as a function of continuum robot configuration, previously proposed models do not consider the effects of gravity on tendon tensions. In this paper, we discuss the addition of gravity to a static model previously developed for low-mass tendon-actuated continuum robots. Using the Euler-Lagrange methodology, the potential energy due to gravity is incorporated into the formulation of the equations that describe tendon tensions as a function of robot configuration. Preliminary experimental results reveal the potential of this nonzero-gravity tendon-tension model.

Visualization of microstructural change affected by mechanical stimulation in tendon healing with a novel tensionless model

Since the majority of a tendon’s dry weight is collagen fibers, tendon healing consists mainly of collagen repair and observing three-dimensional networks of collagen fibers with scanning electron microscopy (SEM) is optimal for investigating this process. In this report, a cell-maceration/SEM method was used to investigate extrasynovial tendon (unwrapped tendon in synovial tissue such as the tendon sheath) healing of an injured Achilles tendon in a rat model. In addition, since mechanical stimulation is important for tendon healing, a novel, tensionless, rat lower leg tendon injury model was established and verified by visualizing the structural change of collagen fibers under tensionless conditions by SEM. This new model was created by transplanting the leg of a rat with a tendon laceration to the back, removing mechanical stimulation. We then compared the process of tendon healing with and without tension using SEM. Under tension, collagen at the tendon stump shows axial alignment and repair that subsequently demarcates the paratenon (connective tissue on the surface of an extrasynovial tendon) border. In contrast, under tensionless conditions, the collagen remains randomly arranged. Our findings demonstrate that mechanical stimulation contributes to axial arrangement and reinforces the importance of tendon tension in wound healing.

Errors in Predicting Muscle Fiber Lengths from Joint Kinematics Point to the Need to Include Tendon Tension in Computational Neuromuscular Models

Accurate predictions of tendon forces must consider musculotendon mechanics; specifically muscle fiber lengths and velocities. These are either predicted explicitly by simulating musculoskeletal dynamics or approximated from measured limb kinematics. The latter is complicated by the fact that tendon lengths and pennation angles vary with both limb kinematics and tendon tension. We now derive the error in kinematically-approximated muscle fiber lengths as a general equation of muscle geometry and tendon tension. This enables researchers to objectively evaluate this error’s significance—which can reach ~ 80% of the optimal muscle fiber length—with respect to the scientific or clinical question being asked. Although this equation provides a detailed functional relationship between muscle fiber lengths, joint kinematics and tendon tension, the parameters used to characterize musculotendon architecture are subject- and muscle-specific. This parametric uncertainty limits the accuracy of any generic musculoskeletal model that hopes to explain subject-specific phenomena. Nevertheless, the existence of such a functional relationship has profound implications to biological proprioception. These results strongly suggest that tendon tension information (from Golgi tendon organs) is likely integrated with muscle fiber length information (from muscle spindles) at the spinal cord to produce useful estimates of limb configuration to enable effective control of movement.

Impact of High Order Wave Loads on a 10 MW Tension-Leg Platform Floating Wind Turbine at Different Tendon Inclination Angles

Floating wind technology is being developed rapidly with the aim of harvesting high-energy wind resources in medium and deep water areas, unreachable using fixed bottom solutions. Given the complexity of these systems, the interactions between the structure and incident hydro-aerodynamic forces need to be well understood. While numerous solutions are being explored, an optimal design is yet to be established within the industry. This study explores the effects of tendon inclination on the dynamic behaviour of a 10MW tension-leg platform (TLP) floating offshore wind turbine (FOWT), and the interaction of different design solutions with higher-order hydrodynamic loading. The model was subject to an extreme sea state in order to capture second and third-order wave effects, and the nonlinear waves were generated via the high-order spectral (HOS) method. The analysis was performed using the hydrodynamic engineering tool CALHYPSO, in-house developed by EDF Lab. Second and third order inertial hydrodynamic loads were included in the time-domain simulations in order to capture low frequency loads and ringing effects respectively. Results show that difference-frequency second order effects have a negligible impact on motions and tendon tensions of the analysed floating wind turbine model, while third order terms can significantly enhance the dynamic response of the system to extreme incident waves. While inclined-leg floater configurations presented improved motion and tendon tension responses under linear loading, the inclusion of quadratic and triple-frequency contributions showed that tendon inclination can in fact increase tension variations in the mooring lines when subject to extreme wave climates. This can lead to slacking in the mooring lines being observed more frequently in inclined-leg configurations. The results therefore suggest that neglecting third order effects, as commonly done in industry, can lead to significant underestimations of motion and tendon tension responses of tension-leg platform wind turbines.