Introduction: Frozen shoulder affects 2 to 5% of the population. Patients experience pain and limited shoulder range of motion (ROM) for an average of 30 months. A 50% reduction in internal and external rotation as well as shoulder elevation are used as diagnostic criteria. The aim of this study is to find if the structural changes in glenohumeral ligaments and the coracohumeral ligament described in literature can account for this reduction in ROM.
Method: An existing shoulder model build in OpenSim was used for this study, to which ligaments were added. The attachment sites of the ligaments as found in literature were matched to positions on the bone surfaces in the model. The rest lengths of the ligaments was determined so that the glenohumeral ROM described in literature could be achieved with less than 5% strain on the ligaments. An increase in ligament stiffness of 50% and decrease in length of 20% was associated with frozen shoulder. Four additional models were created with different attachment sites that are within the anatomical variance.
Results: Shortening the ligaments by 20% results in an average loss of achievable poses of 72.6% for each plane of elevation. The average loss of internal and external rotation over all planes of elevation is 46.0% and 49.3% respectively, while the loss of shoulder elevation is only 31.8%. The role of stiffness changes on the achievable ROM is limited, where a 50% stiffness increase results in a 5.5% loss of available ROM. The loss of internal rotation across all models ranges from 13.3% to 51.6%, from 44.5% to 61.1% for external rotation and 9.6% to 31.8% for shoulder elevation.
Conclusion: Stiffness changes are not enough to account for the loss of ROM associated with frozen shoulder. Shortening the ligaments by 20% results in a loss of ROM of close to 50%. Combining stiffness and length changes would result in a loss greater than 50%. Certain ligament configurations are more susceptible to length changes, which could make them more prone to developing frozen shoulder.

In running and cycling, efficiency and its dependency upon frequency have been investigated extensively. In rowing however, there is still much to be uncovered. In literature, metabolic efficiency seems to be unaffected by stroke rate, despite theory suggesting the existence of some optimal frequency. In order to better understand the mechanisms behind rowing technique, more research is needed.

Until recently, rowing research was largely conducted through experiments. However, muscle analysis from experimental data alone is limited to EMG data collected during the experiment. With the rise of musculoskeletal models, not only muscle activations, but also other muscle variables can be simulated from kinematics and external forces, enabling muscle analysis in more detail.

This report describes the complete process of gathering comprehensive data in an experiment, processing and preparing this data for use in musculoskeletal simulation, and then using this data to validate the model and generate results. Forces, motion, EMG and breathing gas data are collected with three main goals in mind; to extensively validate the musculoskeletal model, to evaluate muscle contributions to total work across different stroke rates and power outputs, and to assess the effects of stroke rate and power output on muscle contributions and metabolic efficiency.

The experimental results, as well as modeling and simulation outputs, in this report, are in accordance with values reported in literature. Additionally, errors in marker tracking and residual and reserve forces and torques have been found to be within acceptable limits, though it should be noted that errors in the upper body are higher than in the legs. Nevertheless, the model is deemed to be valid for the application of ergometer rowing.

Throughout stroke rates and power outputs, no statistically significant effect of stroke rate or power output on muscle contributions have been found. Additionally, the effects of stroke rate and power output on metabolic efficiency are deemed insignificant. Across all stroke rates and outputs, the quadriceps, and more specifically the Vastus Medialis, have been identified as the largest contributors to total work at the muscle level. At the joint level however, the hips are the main contributors, despite the Vasti only acting at the knee. This illustrates that work is exchanged between joints by tendinous action from biarticular muscles such as the hamstrings, an effect also known as Lombard's paradox.

Many are affected by musculoskeletal disorders, for example rotator cuff tears or knee osteoarthritis. Disorders affect muscle coordination, which can negatively impact treatment of these disorders. In order to offer personalized treatment options, muscle recruitment needs to be predicted accurately in these patients. Musculoskeletal modeling can be used to estimate muscle coordination non-invasively, but established methods like Static Optimization (SO) underestimate co-contraction of muscles. Predicting co-contraction is important, because it is associated with musculoskeletal disorders through joint stability, joint stiffness and compensatory muscle control strategies, e.g. to minimize pain. State-of-the-art methods used to predict co-contraction include assistance from electromyography (EMG) measurements or joint stability constraints. The existing Rapid Muscle Redundancy (RMR) solver utilizes a glenohumeral stability constraint to predict co-contraction in the deep rotator cuff muscles. The addition of EMG-assistance to this solver, might improve it's capability to predict co-contraction in superficial muscles as well.

Therefore, the main goal of this thesis is to equip the existing RMR solver with an EMG-assisted cost function. Along with this extension, the RMR solver is revised to be class-based in order to improve modularity, scalability and user interaction. The new implementation is verified and is able to utilize EMG-tracking alongside the glenohumeral joint stability constraint.

Because EMG is difficult to normalize reliably and is less practical to acquire outside lab environments, this study focuses on evaluating the effect of providing one EMG signal at the time. Thus, the second goal involves investigating the effect of tracking single EMG signals on muscle activation using as shoulder model and data. It is revealed that the impact of EMG-tracking single EMG signals on muscle recruitment is minimal, with Mean Absolute Error (MAE) and Zero-normalized Cross-Correlation (ZNCC) changes that are generally less than 0.005 and 0.03, respectively.

Since the role of muscle coordination on joint loading is crucial in pathologies like knee osteoarthritis, the third goal is to evaluate the effect of EMG-tracking on knee joint contact force (JCF). To this end estimations are performed on gait data, which includes in-vivo knee implant reaction forces.
EMG-tracking tasks generally cause higher knee JCF during stance. Tracking the most anterior element of the Gluteus Medius improves knee JCF accuracy, with respect to ground truth in-vivo data, the most on average over three subjects and multiple trials, with the Root Mean Squared Error (RMSE) decreasing with 8% and the Zero-Normalized Cross-Correlation (ZNCC) increasing with 4%, although unreliable EMG-normalization raises questions regarding the legitimacy of these results.
The potential of EMG-tracking is especially demonstrated with tracking the Rectus Femoris muscle in challenge 6, in which more co-contractionis present. Through an increase of the the Rectus Femoris activity as well as its antagonist, Semitendinosus, the knee load estimation is increased at the first peak during stance, which was previously underestimated without EMG-assistance.
However, compared to results from SO and CEINMS, our EMG-assisted RMR solver only achieves lower RMSE in 2 out of 14 trials, although the average difference is only 0.03 body weight.

While the new implementation has not proved itself definitively, it seems promising in predicting co-contraction using EMG-assistance from few EMG signals. In the future this or similar methods might play a crucial role in investigating changes in muscle coordination due to rotator cuff tears, knee osteoarthritis or other musculoskeletal pathologies. This can contribute to more personalized prevention, intervention or rehabilitation.

Spasticity can have a detrimental effect on gait, yet its causal link to gait abnormalities remains unclear. One approach to evaluate the impact of spasticity is to use predictive gait simulations, and another approach is to rely on expert consensus regarding how muscle-specific spasticity affects gait. Currently, neither is fully validated, and it is unclear how these two approaches align. We aimed to compare kinematic gait features according to expert consensus with kinematic effects of muscle-specific spasticity using predictive gait simulations, exploring three modelling approaches: velocity-based, force-based, and a combination of both. We employed a predictive simulation framework with a planar neuromusculoskeletal model, driven by reflex-based control. Spasticity was modelled in four muscles, the soleus, gastrocnemius, rectus femoris, and hamstrings, and individually introduced into our healthy model by increasing reflex gains of the specific feedback mechanism(s). Resulting kinematics were compared with the reference healthy simulated kinematics and checked for the occurrence of expert-based gait features. Overall, 15 out of 21 gait features were present in our simulated kinematics, suggesting good alignment with expert consensus. Kinematic gait features not described by expert consensus were found in hamstrings and rectus femoris simulations. Our combined velocity- and force-based spasticity model showed the best alignment with expert consensus compared to velocity-based and force-based models. This study shows how predictive simulations align with current clinical belief and contribute to a better understanding of the mechanisms underlying spasticity.

3D human motion capture can provide unique information on joint angles and position, and body segment lengths. It has many potential medical applications, including diagnostics, disease monitoring, and clinical decision making. Markerless motion capture makes motion capture accessible and has the potential to objectify motion analysis, and possibly reduce bias. However, neural networks for markerless motion capture can become biased by using unrepresentative training data. Mainstream datasets used for training do not provide detailed information about their subjects characteristics, like skin tone and clothing types per video, making it difficult to study their diversity. This study uses the OpenSim Driven Animated Human (ODAH) pipeline to create three datasets with increasingly diverse subjects and environments and assess the bias and generalization of a biomechanics-aware neural network. The Mean Per Joint Absolute Error (MPJAE), Procrustes Aligned Mean Per Joint Position Error (PA-MPJPE), and absolute relative scale error are determined per video and statistically analyzed together with the video's variables. Significant differences were found in the MPJAE and PA-MPJPE for videos with different motion types, subjects, skin tones, and blinds colors, and significant correlations were found with weight and BMI. The scale error differed significantly based on the subject, skin tone, environmental colors, weight, and height. The model performed significantly better on seen skin tones and environments than unseen skin tones and variables. This study has shown that it cannot be assumed that a neural network’s performance will generalize to different skin tones and different environmental colors, because this will result in greater errors. In future research, the created test data should be expanded to retrain the network and determine the effect of increasing training data diversity.

Musculoskeletal modeling has the opportunity to improve the rehabilitation process after a shoulder injury by monitoring the muscle activations and tendon strain during rehabilitation exercises. The PTbot project combines a robot arm and a Thoracoscapular Shoulder Model (TSM) in order to track the shoulder movement and give a physiotherapist insight on the what happens in a human body, but the performance of the TSM is not yet analyzed. This paper aims to get a better understanding of the TSM by testing its performance in a wide range of motion.

With an existing isometric dataset, the TSM is positioned in five different shoulder configurations. First, the model is scaled in four steps. With the help of the muscle moment arms an normalized fiber length of this scaled model, improvements have been made to the latissimus dorsi. This improved model is scaled a second time whereafter an RMR solver is run. This RMR solver computed the muscle activation levels and joint torque residuals of the TSM. The RMR solver is first run with experimental external forces applied on the hand, in order to validate the TSM by comparing the muscle activations to EMG data. After that, the RMR solver is run with artificial external forces, in order to understand which muscles are activated during push and pull exercises.

The Mean Average Error (MAE) between the muscle activation levels and experimental EMG measurements has been calculated for the validation of the TSM. They show that the TSM performs well when the shoulder planar elevation and shoulder elevation angles, but the performance decreases when these angles increase. The active fiber forces have been calculated and the muscles with the major contribution have been identified for pulling and pushing. The Infraspinatus, rapezius and deltoid play a major role for pushing exercises and the subscapularis, teres major and biceps for pulling exercises. The total muscle force needed for pushing is higher than for pulling.

Exo devices allows users to decrease muscle effort by offloading it to the exo device. Compliance in the physical human-robot interface, originating from flexible cuffs and soft body tissue, can affect the forces transferred. The design of the cuff can alter the value of the pHEI compliance, but it is not yet known to what extent changes in pHEI compliance affect the muscle activation. Due to human variances and the difficulty of measuring objective performance metrics, it is difficult to experimentally pin point cause and effect. Therefore a novel model based method was used which combined experimentally obtained pHEI compliance data with musculoskeletal models to simulate the effect of pHEI compliance on muscle activation. A case study using this method was performed on a subject wearing a passive shoulder exotendon suit. Results indicate a large effect of cuff design on stiffness, damping and cuff migration values. Furthermore, optimising a rigid model and adding compliance afterwards resulted in an increase of total normalised muscle cost. Results from this compliant simulation let to results more closely representing EMG data from another study. Finally, optimising a compliant model results in a total normalised muscle cost equal to that of the rigid case, increased robustness of the exo to configuration errors and increased comfort. This indicates that compliance does not have a detrimental effect on optimised performance when taken into consideration during optimisation.

Background: Computational musculoskeletal models are valuable for understanding finger biomechanics and aiding in medical applications such as postoperative therapy and surgical planning. These models simulate various scenarios and predict outcomes by representing the biomechanical system through equations of motion. However, the complex mechanics of the human finger, particularly the role of the Lateral Bands in coordinated joint flexion, have primarily been studied in static simulations. Their role in dynamic simulations, especially with the inclusion of other key tendons and ligaments, has not been researched.
Aim: This study aims to develop a musculoskeletal model to enhance our understanding of the contribution of Lateral Bands to coordinated finger flexion, by combining insights from previous research to address existing gaps in finger biomechanics. The model will replicate the anatomical joints and Lateral Band structures of an anthropomorphic mechanical finger, offering a simplified design to reduce simulation complexity and enhance verification reliability.
Method: An adapted anthropomorphic mechanical finger was created based on previous work, scaled down to reduce weight and required actuation force. The simulation model was developed in OpenSim, using bone geometries from CAD files and direct measurement attachment location of tendons and ligaments from the mechanical finger for improved accuracy. Model verification involved comparing the model’s behavior with the mechanical finger through experiments focusing on finger flexion under external deep flexor tendon load, using motion capture data, inverse kinematics (IK), and forward dynamics (FD). Results The simulation model accurately replicated the mechanical finger’s joint kinematics, with an average squared error ranging from 1.54e−3 mm to 9.55e−3 mm and an average marker RMSE of ±0.65 mm. Tendon displacement and moment arm were verified with average errors ±1 mm and ±1.23 mm, respectively, indicating that the model closely matches the mechanical finger’s tendon force and joint torque relationships. The model demonstrated that Lateral Bands are essential for interphalangeal joint (IPJ) coupling, as they adjust their tension by becoming taut or slack during finger flexion to distribute forces effectively. Without these Lateral Bands, finger flexion leads to unrealistic joint movements in mechanical finger models. Additionally, the Lateral Bands play a significant role in generating tension in the deep flexor tendon during IPJ coupling.
Conclusion: This study demonstrated the role of Lateral Bands during dynamic finger flexion by incorporating anatomically-based structures and replicating their behavior. The findings provide deeper insights into their function and set a new foundation for future research. Additional studies could further explore how Lateral Bands influence force distribution and joint mechanics, potentially leading to better understanding and treatment of finger injuries and disorders.

Background : There have been several advancements in modelling the behaviour of the finger and its internal anatomical structures and in the development of anthropomorphic fingers. Most studies that include ligaments focus on force distribution in the tendon-ligament network for a particular task. In the design of anthropomorphic fingers, especially in two recent projects, ligament attachments have been decided through iterative trials and the the tendon network has been used to control the range of motion of the finger. However, the collateral ligaments also play a major role in limiting the range of motion of the finger in addition to contributing to medial-lateral joint stability.
Aim : The aim of this project is to develop an algorithm to automatically compute coordinates for collateral ligament attachment sites for the Distal Inter-Phalangeal (DIP) joint of the finger, given a required range of motion in terms of maximum flexion and extension angles.
Method : The DIP joint was described in two dimensions (in the sagittal plane) with one degree of freedom (flexion-extension). Constraints were formulated based on geometry and observed and reported ligament length ranges. The model formulation was also informed by insights from the dissection of three preserved human index fingers. The algorithm named ’Auto-LINC’ (Automating Ligament Insertion Coordinates) involved two stages. The first was a combinatorial approach used to establish the best possible combination of attachment sites out of a set of feasible solutions. This was used to seed the second stage - a continuous domain approach used for local optimisation. The algorithm was designed in such a way that the two stages can be used separately or in combination depending on the application.
Results : The attachment sites computed using the combined approach were used to add ligaments to an existing OpenSim simulation model of an anthropomorphic DIP joint to verify that the required range of motion was achieved. The process was illustrated with an example in which 99.6% of the required range of motion was achieved using ligaments with attachment sites computed by Auto-LINC.
Discussion : The use of an algorithm with geometric constraints and objective function yielded feasible solutions in a short time. By adapting the constraints to suit differences in geometry, Auto-LINC
can also be used to predict collateral ligament attachment sites for other flexion-extension joints such as the Proximal Inter-Phalangeal (PIP) joint. Auto-LINC has potential for use in the design of anthropomorphic fingers as it automates the process of deciding ligament attachment sites. It also has potential for use in guiding robot-assisted reconstruction surgery. It can be made more detailed and versatile by adding material properties and constitutive equations and by extension into three dimensions.
Conclusion : Auto-LINC is already usable in its current form to compute form based on function, and has potential for use in automating anthropomorphic design and guiding reconstruction surgery. In
contrast to other published research, Auto-LINC computes ligament attachment sites based on range of motion - in other words, it focuses on moving from function to form rather than form to function.

The ability to simulate the impact on performance of military body-borne loads enables effective analysis and optimisation of load and equipment configurations for military personnel performance. Additionally, these simulations can reduce research & development costs of new equipment by analysing its impact in an early stage of development. While research has been done into the effects of body borne loads on kinematics, ground reaction forces (GRF) and metabolic cost of transport, the accuracy and reliability of simulations are not clear yet.
This study set out to predict loaded gait kinematics and GRFs and estimate metabolic cost of transport for gait at 1.5 m/s, carrying different types of military relevant body-borne loads, to evaluate the applied methods for implementation in a future load configuration optimisation tool.
The kinematic/GRF prediction was performed by forward dynamics simulations in SCONE/hyfydy, using a planar musculoskeletal model and a 2 gait-state controller, simulating 15 solutions for each load condition. While only 60% of experimentally measured differences between load conditions were correctly predicted, the expected differences from literature were all correctly predicted. It was assessed that improving the number of gait states and the number of optimisations per load condition is expected to improve these results.
The metabolic cost of transport (mCoT) estimation was performed by the Computed Muscle Control algorithm of OpenSim, using experimental kinematics and ground reaction forces. However, small compounding errors in experimental data and data processing prevented accurate mCoT estimations. Although the applied kinematic/GRF and mCoT simulation methods could not be validated yet, based on these results, the study as a whole does show promise for the continued development of these models and their future implementation for loaded gait performance optimisation.

Introduction: The BMX start is crucial for race performance, often measured by the time to the kink at 3.15 m from the start.
Objective: This study aims to optimize the BMX SX gate start using predictive optimal control techniques, focusing on the effects of maximal crank torque and reaction time on performance.
Method: Two models were used: The ‘upper extremities’ model analysed varying crank torques (250 Nm to 350 Nm) and reaction times (0.14 s, 0.16 s, 0.18 s).
The ‘two legs’ model was assessed under a single condition to better reproduce crank torque and track experimentally measured kinematics.
Results: Higher crank torques led to more forward initial positions, reduced recoil, and increased final velocities. The velocity of the start gate was a limiting factor initially, with timing and technique being crucial until the gate is halfway open. Reaction time variations showed minor effects on performance, and no strategy adaptation was needed within the tested range. The ‘two legs’ model accurately tracked experimental kinematics with low RMSE values. The predictive simulation with the ‘two legs’ model showed an improvement in kink time. The kink time for the predictive optimal control solution was 1.15 s compared to 1.23 s in the experimental trial.
Conclusion: This framework for researching the BMX start using predictive optimal control offers a systematic basis for future research. Insights can improve training strategies focusing on technique, timing, and initial start position. Future research could explore the effects of leg strength, hip range of motion, and gear ratios or crank lengths on performance.

Background: Abnormal patellofemoral joint loading patterns may cause patellofemoral pain and may be caused by patellar maltracking. Treatment is aimed at restoring normative patellofemoral kinematics. For this, an accurate estimation of the patellofemoral kinematics is necessary. While 4-dimensional computed tomography (4D-CT) can provide this by capturing bone geometries, this equipment is not commonly available in clinical settings and requires radiation. Optical Motion Capture (OMC) with a grid of markers overlying the bone has shown promising results for measuring scapular and patellofemoral kinematics.
Objective: The aim of this study is to develop and validate a new method to estimate patellofemoral kinematics from OMC by applying a shape fitting algorithm to a knee marker grid.
Methods: Five participants were equipped with a knee marker grid and performed a prone extension-flexion movement during which kinematics were measured using OMC and 4D-CT. Patellofemoral kinematics were estimated using three methods: Geometry-based 4D-CT, Grid-based 4D-CT, and Grid-based OMC. In Geometry-based 4D-CT, PF kinematics were obtained from bone geometries. In both Grid-based methods, a shape fitting algorithm using an Iterative Closest Point algorithm estimated the patellofemoral kinematics from the locations of the grid markers. For validation, both 4D-CT methods were compared, as well as both Grid-based methods. To quantify the validity of the new method, root mean square errors (RMSEs) of the differences between the methods and Spearman correlation coefficients were computed.
Results: For Geometry-based 4D-CT vs. Grid-based 4D-CT, the RMSEs of the differences were 7° for flexion and 3 to 10 mm for all translations, combined with very high correlations. The RMSEs for tilt and spin were over 18°, combined with low and moderate negative correlations. For Grid-based 4D-CT vs. Grid-based OMC, RMSEs were smallest for ML translation (4 mm) and spin and tilt (< 6°), but these were paired with moderate correlations. On the other hand, the higher RMSEs for flexion (9°) and AP and SI translation (> 10 mm) were paired with moderate to very high correlations.
Conclusion: Comparing the 4D-CT methods indicated that spin and tilt could not be measured accurately using the marker grid, which are both clinical relevant. Comparing the Grid-based methods showed large differences between OMC and 4D-CT, probably introduced by the accuracy of the measurement system. Overall, the shape fitting algorithm thus seems to be able to estimate PF flexion and all three PF translations from a marker grid, but applying it to OMC data is not valid yet.

Manual wheelchair users, especially those with spinal cord injuries, often suffer from shoulder overuse and pain. Understanding the specific role of the muscles involved in wheelchair propulsion is essential to prevent these complications. Yet, there is uncertainty about which muscles are primarily recruited, and their contribution to propulsion power. For example, some studies have suggested that the rotator cuff muscles are the primary contributors to wheelchair propulsion. Several studies have explored muscle activation patterns using musculoskeletal models, but often omit glenohumeral stability and key muscles like the trapezius, serratus anterior, and rhomboideus. Investigating muscle work provides insights into the actual mechanical output, offering a more accurate assessment of efficiency by considering factors like muscle force and movement distance.

The aim of this study was to examine individual muscle contributions to mechanical work during the push and recovery phases of wheelchair propulsion using a musculoskeletal model and to identify the role of rotator cuff muscles. The thoracoscapular shoulder model featuring accurate scapula kinematics, inclusive of the rhomboideus, serratus anterior, and trapezius muscles, was employed. In a previous experiment electromyography, kinematic and pushrim forces of 5 persons with paraplegia were collected. The filtered pushrim forces and the kinematic data as well as the scaled musculoskeletal model were inputs to the rapid muscle redundancy solver, that enforced glenohumeral joint stability while estimating muscle activations and muscle power. Muscle work was integrated from the estimated muscle power and normalized by the total mechanical work. During the push phase, significant contributions came from the pectoralis major, anterior deltoid, infraspinatus, serratus anterior, triceps brachii, and biceps brachii. In contrast, the recovery phase primarily involved the teres major, subscapularis, trapezius, posterior deltoid, middle deltoid, and rhomboideus. The contribution of rotator cuff muscles to propulsion was noted but to a lesser extent than previous reports, with no contribution from the supraspinatus. Surprisingly, the teres major showed high work values, possibly due to insufficient activation of the latissimus dorsi. Our findings support the importance of incorporating muscles like the serratus anterior, trapezius, and rhomboids into musculoskeletal models. Furthermore, the contribution of reserve actuators to total work generation was below the threshold of 5%. The comparison between estimated muscle activations and measured electromyographic activations generally showed excellent to good magnitude matching. This study's outcomes can aid in designing ergonomic wheelchairs and guide the development of tailored rehabilitation and training methods to decrease upper extremity stress in wheelchair users.

Inertial Measurement Units (IMUs) have become increasingly popular human motion estimation due to their portability, self-contained features, and cost-effectiveness compared to marker-based sensing systems which rely on external cameras to observe the position of the markers. Over the years, many studies have proposed various models and algorithms based on IMUs and the kinematics of the human body for motion estimation, neglecting the fact that IMUs measure acceleration, which can be directly related to joint torque with known inertial parameters. In turn, by estimating joint torque and incorporating kinetics into the model, it becomes possible to address a long-standing problem in the field of biomechanics: the marker-based sensing system’s inability to provide a reliable estimation of kinetics due to the need for numerical differentiation. In a previous study, a kinetics model was proposed for estimating the motion but validated only on a robotic arm. In our study, we further explore the performance of using IMUs to estimate wheelchair user motion data based on Extended Kalman Filter (EKF) and the kinetics model, with marker-based Inverse Kinematics (IK)/Inverse Dynamics (ID) as the benchmark.

Compared to Marker-based IK, the method leveraging kinetics achieves a Root Mean Squared Difference (RMSD) below 16◦ for three out of four tasks across all joints throughout the trial. After analyzing the only task with degradation in estimation, we conclude that the erroneous IMUs measurements results from Soft Tissue Artefacts (STA) is the most likely reason. For joint torque estimation, the RMSD for joint torque estimation is below 3.05Nm for the tasks less affected STA. Through fine-tuning the EKF, we can achieve fast and responsive estimation results without being affected by numerical differentiation, enabling us to capture both sudden and subtle changes in joint torque estimation. The kinetics model performs better than the kinematics-based model in estimating both kinematics and kinetics and also reduced drifting behavior. Compared to OpenSense, which depends on magnetometer measurements, kinetics model estimation shows comparable kinematics estimation accuracy while excluding
the use of heading information. The results show that including the kinetics model for human motion estimation can improve estimation accuracy and robustness encouraging further studies to include kinetics for human motion estimation.

Children with cerebral palsy (CP) commonly have bony deformities of the foot, which lead to pain and gait problems. One of the causes of such a deformity is an imbalance in muscle forces around the foot. In turn, the bony deformity can also alter muscle function, due to, for example, altered muscle moment arm lengths. In this study, the altered muscle function due to hindfoot varus deformities in children with CP was investigated using musculoskeletal models. The first aim of this study was to create personalized musculoskeletal models of the foot based on weight-bearing computed tomography (WBCT) data. Secondly, we applied the model to get insight in how joint axis orientations and muscle moment arms are altered in varus feet, and how this leads to differences in muscle function during gait.
Models were created in OpenSim for seven children with a hindfoot varus deformity due to CP, and four adults with neutral feet. Each model was based on WBCT scans and included five degrees of freedom in the foot (talocrural, subtalar, Chopart, Lisfranc and metatarsophalangeal joints). The orientations of the foot joint axes and moment arms of the extrinsic foot muscles were calculated. Subsequently, the models were combined with motion capture and ground reaction force data to calculate muscle activations during gait.
The joint axis orientations showed greater variability in the group of children with CP compared to the control group; in most subjects, changes in axis orientation were observed that may lead to a more rigid foot. Furthermore, the dorsiflexion moment arm of the tibialis anterior decreased, while the inversion moment arm increased; thus, the tibialis anterior became an even more effective invertor when a varus deformity of the foot was present. On the other hand, the eversion moment arms of the peroneal muscles tended to become smaller, meaning they would be less effective in resisting the varus deformity. Static Optimization results showed decreased activity in the tibialis muscles, and increased activity in the peroneal muscles. This increased activity might be necessary due to the smaller moment arms, and/or to stabilize the ankle.
This is the first study in which a musculoskeletal foot model was developed based on personalized bone data of feet with bony deformities in weight-bearing conditions. Distinct changes were shown in muscle function when a varus deformity is present, which might lead to progression of the deformity. This contributes to a better understanding of the altered muscle function due to foot deformities, which may eventually contribute to improvement of treatment to prevent the progression of foot deformities.

Biomechanics studies the underlying mechanisms between body movements and forces. Accurate motion data is crucial for the biomechanics. Currently, marker-based motion capture systems are often used by researchers to record motion data. Marker-based motion capture systems are not widely adopted due to its drawbacks in terms of financial and time costs, portability, etc. Video-based motion capture systems can record motions using videos collected by webcams, cameras, and smartphones as the input and then estimate human motions from those videos. A simpler setup makes video-based motion capture technology more accessible for widespread use. However, existing motion capture datasets commonly lack of biomechanically accurate annotation, resulting in a deficiency in the biomechanical accuracy of exsisting video-based motion capture methods. In the biomechanics community, there are a lot of validated and biomechanically accurate models and motion data; however, corresponding video data is lacking. We can construct human-like appearance based on these data and generate a synthetic human motion video dataset using 3D graphic software. In this thesis, we purposed a pipeline that can generate synthetic human motion videos. The pipeline takes subject-specific OpenSim model and motion as input and uses SMPL-X model to generate human-like appearance. We validated the synthetic data generated by our pipeline and demonstrated the biomechanical reliability of the pipeline. Using this pipeline, we created synthetic dataset ODAH with biomechanically accurate annotations for neural network training.

Shoulder exoskeleton is a popular solution to work-related shoulder disorders and muscle fatigue. With a wide range of exoskeletons designed, a comprehensive report on how the use of shoulder exoskeletons changes shoulder biomechanics is still missing. In this project, the impact of exoskeletons on shoulder biomechanics was investigated with the musculoskeletal simulation OpenSim. This study proposed a "human-in-the-loop" optimization-based design tool for shoulder exoskeletons. This design tool incorporates the predicted biomechanical effects of a shoulder exoskeleton from musculoskeletal simulations into design considerations. This design tool was validated with a case study designing a shoulder exoskeleton based on a compliant beam and testing the design in the musculoskeletal simulation and experiments.

The exoskeleton design tool is a coupling of finite element analysis and OpenSim. OpenSim calculates the deformation of the exoskeleton with human motion, and the finite element analysis calculates the force exerted from the exoskeleton upon deformation. Then OpenSim computes muscle activities under the external force from the exoskeleton. By merging muscle activities and the resultant glenohumeral joint reaction force to an objective function, the optimization-based design loop is closed by looking for the best objective value iteratively.

Several exoskeletons were designed by the new design tool to assist different types of tasks. The design tool exhibited good ability in finding optimal solutions for a range of design choices and design requirements. Simulated tests of designed exoskeletons showed significant effects on reducing muscle activities and good robustness in resisting the influence of perturbed motions in arm-elevated tasks. An exoskeleton was selected to be tested with an experiment set up in the same way as the simulated test. Experiment results supported the performance of the exoskeleton predicted in the simulated test.

This project established a method to comprehensively predict the effect of an exoskeleton on shoulder biomechanics and provided a more comprehensive understanding of biomechanical effects of shoulder exoskeletons. This facilitated the “human-in-the-loop” design process of shoulder exoskeletons which could greatly save money and time investments into prototyping, testing, and validation.

The objective of this study is to simulate lumbar stabilization in autonomous vehicles through the utility of OpenSim. Considering sufficient isometric strength is required for dynamic analysis, validity of the full-body model developed by Schmid et al. [43] in isometric conditions is examined first. This encompasses the establishment of maximal isometric force exterted by the trunk in sagittal and frontal plane. The model is solved using an inverse-dynamics based static optimization and the model-predicted strength was taken to be the highest external force for which the Static Optimization Tool successfully solved [46]. Finally, isometric model conditions are validated against literature. Predicted strength correlated within 1 standard deviation (SD) of the in vivo measurements and therefore, yielded realistic results in forward, rearward and lateral direction.

In this work, we propose a method for monitoring and management of rotator-cuff tendon strains in human-robot collaborative physical therapy for rotator cuff rehabilitation. The proposed approach integrates a complex offline biomechanical model with a collaborative, industrial robot arm and an impedance controller. The model is used for computing rotator-cuff tendon strain as a function of human shoulder configuration, muscle activation and external forces. This subject- and injury-specific data is stored in \textit{strain maps}, which represent the relationship between the strains and shoulder DoFs. In our previous work, we implemented strain maps to preplan minimal strain, safe trajectories using two shoulder DoFs, and used the corresponding robot-mediated movement for passive trajectory following for healthy subjects. This work expands on that by implementing two novel functionalities: 1) patient-led movement, and 2) adding the third shoulder DoF and the corresponding control complexities, while still controlling for safe rotator-cuff tendon strains. For patient-led movement, we precomputed unsafe zones for each strain map by clustering and fitting ellipses to the clusters. These unsafe areas with increased risk of (re-)injury are then used to set the impedance control parameters and reference pose for real-time biomechanical safety control. By linearly interpolating between strain maps, smooth and safe movement of the third shoulder DoF is added. The resulting robot control torques guide the patient away from unsafe, high strain shoulder poses in real-time during patient-led movement. The proposed method has the potential to improve the safety, Range of Motion, and muscle activity that the patients receive through robot-mediated physical therapy. The main advantage of this approach is that the patient is free to use and explore their full shoulder RoM, while the robot controls and manages biomechanical safety in real-time. To validate the proposed method, we performed two experiments showcasing two novel functionalities, and a third experiment as proof-of-concept displaying the full method.