The inspiration is anatomical. The human neck consists of seven vertebrae whose coordinated action with joints, ligaments and muscles produces an extraordinary repertoire of movement: flexion, extension, lateral bending and rotation. Existing humanoid robots have struggled to copy this. Honda’s ASIMO and Honda–AIST’s HRP-2 carry serial neck structures with just two degrees of freedom, while later machines such as Albert Hubo and iCub added a yaw joint for more complex head motion. Parallel neck designs, which use multiple actuators working against a common platform, offer higher rigidity and lower inertia, but the authors argue they have consistently lacked the compliance needed to simulate the bending motion of a real cervical spine.

The new robot attacks that gap with a five-part architecture: a moving platform, a fixed platform, a six-bar spring system, a rotating mechanism and an external flexible support system, all driven by three lightweight cables. The six-bar spring system — three sets of crossed spring push rods distributed at 120-degree intervals between the two platforms — plays the role of the cervical vertebrae, providing structural stability while allowing controlled bending compliance. The external flexible support system mimics the neck muscles, and a motor-driven rotating mechanism adjusts that system’s radius across an angle range of 0 to 60 degrees, changing the cable geometry and thereby the robot’s overall stiffness in real time.

Controlling such a hybrid machine is the hard part, because rigid and flexible elements deform together in ways that are difficult to predict. The team’s solution is a mathematical mapping they call function G, which links the robot’s configuration space — the bending angle and azimuth angle of the platform — to its actuation space, the length changes of the three driving cables. Cable length variations are constrained within ±40 millimetres. Because the mapping is invertible, the researchers also derived a companion function A that recovers the robot’s posture from measured cable lengths, allowing force–position control in which motors adjust cable tension to hold or change a pose precisely.

On top of the kinematics sits a full dynamics model. The team built a static analysis using torque and force balance based on Newton’s laws and the conservation of angular momentum, calculating how cable tension distributes against the resistive torques of the spring system and the flexible support in any bent posture. They then extended this into dynamic territory with Euler–Lagrange equations, incorporating inertial forces, Coriolis and centrifugal effects, and the elastic potential energy stored in both spring systems. The result is a coupled rigid–flexible dynamic model in which the driving torque of the cables balances the resistance torque generated by the deforming flexible structure — the mathematical analogue of muscles straining against bone.

The stiffness trick is the study’s centrepiece. When the rotating mechanism turns, it slides the attachment radius of the external flexible support system outward or inward, altering the pre-tension and effective stiffness of the whole assembly without changing the cable lengths that hold the pose. In a fixed posture, the robot can therefore sit still while its resistance to external disturbance rises or falls — exactly the stabilising function human neck muscles perform when the head is jostled. Simulations across four different values of the adjustable radius show overall stiffness climbing smoothly as the rotation angle increases, with the total stiffness combining contributions from the six spring rods, the three support springs and the cables themselves.

The quantitative results are grounded in human biomechanics. The maximum bending angle was set at 30 degrees, reflecting the normal flexion range of the human neck reduced by roughly 30 percent to account for the restriction caused by wearing heavy protective equipment. Under cable tensions ranging from 0 to 80 newtons, the robot reached that angle while cable tension and actuation curves followed clean trigonometric trajectories across the workspace. The team also mapped the robot’s task space in three dimensions, showing that cable trajectories remain within the platform’s motion envelope throughout — an important sanity check for a machine intended eventually to carry a head-like load.

The authors are candid about what the design does not yet do. The current stiffness adjustment works by changing the robot’s external dimensions, which may compromise biomimetic fidelity and practical packaging. Friction was neglected in the dynamic calculations, even though the team acknowledges it can measurably degrade motion performance and end-effector accuracy in real hardware; they propose lubricants, reduced contact areas and self-lubricating bushings as mitigations. The model also omits nonlinear cable friction and structural hysteresis, and the stiffness response speed and adjustment accuracy have not yet been verified on a physical prototype. Future work will integrate the mass and volume distribution of a human head into the load profile, refine the bone–muscle coupling with detailed anatomical data, and add force and pose sensors for closed-loop feedback control.

Even so, the study stakes out a clear position in the race toward lifelike humanoid machines. Compared with an earlier cable-driven design with a spring spine by Gao and Song, the new robot expands the usable bending range and more faithfully reproduces the interplay of cervical bones and muscles. Beyond humanoid heads, the authors see the architecture as a reference for cable-driven flexible mechanisms generally — a class of machines that includes surgical manipulators, rehabilitation devices and soft industrial arms. If a robot is going to work alongside people, absorb bumps, carry loads and move with something approaching natural grace, the neck may be the place where that realism is decided, and this machine offers a concrete, quantified step toward it.

The work, published as volume 17, pages 917–933 of Mechanical Sciences under a Creative Commons licence, was supported by a science and technology innovation and entrepreneurship training program for college students at Jilin Agricultural Science and Technology University. Its deeper significance may lie in the design philosophy rather than any single number: instead of treating compliance and precision as competing goals to be traded off, the robot treats them as two faces of one system, coordinated the way evolution coordinated them — bones for structure, muscles for adjustable tension, and a control law that keeps the two in conversation. That principle, validated here in simulation with explicit stiffness figures and workspace maps, gives roboticists a template for building machines that are simultaneously strong, precise and soft enough to be safe around the people they are built to serve.

Subject of Research: A cable-driven parallel robot with variable stiffness and rigid–flexible coupling that replicates human cervical spine motion

Article Title: Design and analysis of a novel cable-driven parallel robot with variable stiffness and rigid–flexible coupling

Article References: Wu, Y., Xiao, X., Wang, H., & Song, D. (2026). Design and analysis of a novel cable-driven parallel robot with variable stiffness and rigid–flexible coupling. Mechanical Sciences, 17(2), 917-933. https://doi.org/10.5194/ms-17-917-2026

Image Credits: AI Generated

DOI: 10.5194/ms-17-917-2026

Keywords: cable-driven parallel robot, variable stiffness, rigid–flexible coupling, cervical spine, biomimetic robotics, humanoid robots, kinematics, dynamics modeling, Euler–Lagrange equations, compliant mechanisms, Mechanical Sciences, stiffness modulation

The subject-wise six-fold cross-validation design is what makes this comparison meaningful. In many biomechanics applications of machine learning, data from a single athlete appear in both training and testing sets, allowing the model to memorise individual quirks such as stroke timing, hand size, and force magnitude. Here, each held-out fold consisted entirely of swimmers the model had never seen, so any generalisation had to come from shared structure in the kinematics-to-force relationship rather than from person-specific memorisation. The nested structure, in which architecture, regularisation, and smoothing choices were made inside each training fold, further protects against the subtle leakage that occurs when hyperparameters are tuned against the test data. Participant-cluster bootstrapping then quantified how sensitive the reported errors were to the particular composition of the twenty-four swimmers, an important safeguard given that a sample of this size, drawn from national and regional sprinters, cannot represent the full spread of competitive swimming.

The failure of between-swimmer correlations, which ranged from −0.36 to 0.27 for peak force, impulse, and mean force, illustrates a phenomenon well known in gait analysis but rarely quantified this explicitly in swimming. A model can track the temporal pattern of a waveform almost perfectly while producing outputs that are essentially the same for everyone, shifted and scaled only slightly. In such a case, the correlation between measured and estimated summary values across individuals hovers near zero, because the estimates lack the variance needed to rank athletes correctly. For a coach deciding which of two swimmers generates more impulse per stroke, this is the failure mode that matters most, and it is invisible in a pooled coefficient of determination. The study’s multi-level evaluation framework, examining the waveform, discrete descriptors, between-swimmer differences, and inter-limb differences separately, offers a template that future wearable-estimation studies in cyclic sports would do well to adopt.

The inter-limb findings are particularly instructive because asymmetry is a genuine clinical and performance concern in swimming. Swimmers often breathe to one side, develop side-dominant strokes, and carry asymmetric loads into the shoulder girdle, so an instrument that could flag meaningful left-right differences without hand-worn hardware would have real value. The estimated symmetry index, however, correlated negatively with the measured one, and the model identified the dominant limb no more often than chance would roughly allow, matching the measured dominant side in only 50 to 58 percent of swimmers. A negative correlation is arguably worse than a zero one, since it suggests the estimates could actively mislead a practitioner about which arm produces more force. Combined with the systematic 32.6-newton underestimation of left-hand peak force, this points to a lateral asymmetry in the model’s learning, possibly related to marker placement conventions, breathing-related trunk rotation, or an imbalance in the effective training examples available for each hand.

The comparison between analysis window definitions reveals how easily evaluation choices can flatter a model. When both hands were modelled in every window, including each hand’s recovery and entry periods, apparent accuracy rose, with left-hand root-mean-square error dropping to 21.6 newtons and correlation climbing to 0.89. But during recovery and entry, the hand is largely out of the propulsive flow, and its force output is low and stereotyped, so including those windows pads the evaluation with easy predictions. The stricter definition, restricting analysis to fingertip entry through exit, is the one that matches how the underwater phase is conventionally defined in swimming biomechanics, and it is the one under which the honest performance figures were obtained. This is a methodological lesson with broad relevance: the apparent accuracy of a kinematics-to-kinetics estimator depends not only on the model but on where in the movement cycle the evaluation is conducted.

The choice of a multilayer perceptron over more fashionable architectures reflects a deliberate emphasis on validation rather than novelty. The network consumed 197 features expressed in a trunk-fixed reference frame, a choice that removes the confounding effect of whole-body orientation and translation, so that the model receives limb posture relative to the torso rather than absolute coordinates in the pool. Comparisons against gradient boosting and coordinate-only variants showed that the multilayer perceptron was broadly comparable to other learned approaches, meaning the ceiling was set by the information content of the kinematics rather than by the choice of learning algorithm. This is an important negative result. When several model families converge on similar error against a simple template baseline, the bottleneck lies in the input data, and further architectural refinement offers diminishing returns until richer inputs, such as hand orientation, hand surface area estimates, or wrist-level inertial signals, are supplied.

The physics of the problem explains why the kinematics-to-force mapping is fundamentally harder in water than on land. Ground reaction force during walking and running emerges from a rigid contact whose mechanics are tightly constrained by whole-body motion, which is why convolutional networks can estimate it from joint angles with normalised errors below roughly thirteen percent in healthy adults. Hand propulsion in front crawl, by contrast, arises from unsteady hydrodynamics: the force depends on hand velocity relative to the water, hand orientation, the effective propelling area, and the behaviour of vortices and shed flow structures, some of which leave only weak traces in marker trajectories. Two swimmers with nearly identical arm kinematics could, in principle, produce different forces if their hand shapes, finger spread, or the flow history around the hand differ. The kinematic variables therefore carry substantial but incomplete information about force, and the residual gap is exactly what the template baseline was able to capture without any individual-specific learning.

The device-on protocol also deserves scrutiny when interpreting the error magnitudes. Because the swimmers wore the eo SwimBETTER handsets while the motion capture was recorded, the kinematics available to the model included any alterations in hand posture or stroke pattern caused by the handsets themselves. This design choice ensures that the force reference and the kinematics come from the same physical execution of the stroke, avoiding synchronisation and condition-mismatch errors, but it also means the validated model is conditioned on instrumented swimming. A model trained this way might perform differently on swimmers not wearing handsets, whose hands and wrists move slightly differently. Conversely, the device-on approach guarantees that the reference force measurements are genuine in-pool recordings rather than laboratory simulations, which strengthens the ecological validity of the force targets even as it narrows the deployment scenario in which the model has been tested.

The participant characteristics frame the generalisation boundary of the current results. The cohort consisted of twenty-four male sprinters with a mean 50-metre season best of about 24 seconds, roughly nine and a half years of training experience, and weekly training volumes around 42 kilometres. Female swimmers, distance specialists, developing age-group athletes, and recreational swimmers all differ in stroke mechanics, hand size, and force magnitude, and none are represented here. Because the model failed to discriminate between swimmers even within this relatively homogeneous group, there is little reason to expect better between-individual performance in a more diverse population without substantial additional data. The sample size, set by the availability of eligible sprinters rather than by a formal power calculation, nonetheless exceeds the ten swimmers used in the original device validation, and the bootstrap uncertainty estimates give a sense of how stable the reported errors are across resampled cohorts.

Looking forward, the study clarifies what kinematics-driven estimation could realistically offer in the near term and what remains out of reach. Recovering the group-typical force-time pattern from marker data alone is now demonstrated under rigorous subject-wise validation, which means kinematic archives could be used to study population-level stroke dynamics, compare conditions within a swimmer, or examine how force patterns change across a race or a training block, provided the questions concern temporal structure rather than individual magnitude. For individual assessment, ranking athletes, quantifying asymmetry, or tracking force production longitudinally in a single athlete, direct measurement with hand-worn devices remains necessary. The most promising extensions would supply the model with information closer to the hydrodynamic determinants of force, such as hand orientation and estimated effective area, or would combine sparse direct force measurements with dense kinematic records so that the model learns to personalise its outputs from a small amount of instrumented data per athlete. Until such hybrid approaches are validated with the same multi-level, subject-wise rigour applied here, hand force in swimming remains a quantity that must be measured, not merely inferred.

Subject of Research: Estimating bilateral hand kinetics from upper-limb kinematics in sprint front crawl: a subject-wise validation of an artificial neural network

Article Title: Estimating bilateral hand kinetics from upper-limb kinematics in sprint front crawl: a subject-wise validation of an artificial neural network

Article References: Gao, T., Qi, S., Li, J., Huang, D., Zheng, K., Li, H., & Gao, J. (2026). Estimating bilateral hand kinetics from upper-limb kinematics in sprint front crawl: a subject-wise validation of an artificial neural network. Frontiers in Bioengineering and Biotechnology, 14, Article 1978998. https://doi.org/10.3389/fbioe.2026.1978998

Image Credits: AI Generated

DOI: 10.3389/fbioe.2026.1978998

Keywords: Estimating, bilateral, hand, kinetics, upper-limb, kinematics, sprint, front, crawl, subject-wise, validation, artificial

The work, published in Frontiers in Artificial Intelligence, addresses one of the most fundamental problems in robotics. Inverse kinematics is the task of determining the joint angles of a manipulator so that its end-effector reaches a specified position and orientation in space. For a standard six-degree-of-freedom arm, the problem has a finite number of solutions. But for a redundant seven-degree-of-freedom manipulator, the same end-effector pose corresponds to a continuous one-dimensional manifold of admissible joint configurations, known as the self-motion manifold. Classical numerical solvers such as damped least squares and the Levenberg-Marquardt algorithm of the Orocos KDL library return only one configuration, and they do so depending on the initial approximation and regularization parameters.

Machine learning approaches have tried to approximate the mapping from end-effector pose to joint angles using neural networks, but most retain a regression-based formulation. The model is trained to predict a single joint vector for a given pose, which fundamentally limits its usefulness for redundant manipulators. When the true solution distribution is multimodal, minimizing the mean squared error leads to averaging across different modes, producing a joint vector that may not belong to the set of kinematically admissible configurations at all. The Ukrainian team, led by Lesia Hentosh with Yaroslav Hladun and Ivan Tsmots, recognized that redundant inverse kinematics requires a distributional formulation rather than a pointwise approximation.

Generative models offer a natural framework for such a formulation, and one of the closest prior approaches is IKFlow, which uses normalizing flows to generate diverse inverse kinematics solutions. But normalizing flows require invertible transformations, which restricts the network architecture and complicates the use of arbitrary residual blocks, normalization layers, and nonlinear transformations. Diffusion models offer a different generative paradigm. They learn to progressively recover data from a noisy state without requiring an explicit invertible mapping between distributions. The team adapted this paradigm to the joint configuration space of a seven-degree-of-freedom xArm manipulator.

The DiffusionIK method operates in three stages. First, a conditional diffusion model with a Residual Multi-Layer Perceptron denoiser generates a set of initial joint configurations in a normalized space. The model is trained using procedurally generated training pairs obtained from the forward kinematics of the manipulator, removing any dependence on a classical inverse kinematics solver. Second, classifier-free guidance strengthens the consistency of the generated configurations with the target end-effector pose, reducing the pre-refinement position residual from 56.3 millimeters to 16.9 millimeters. Third, Denoising Diffusion Implicit Models sampling reduces the number of generation steps, and a differentiable forward-kinematics refinement stage then reduces the task-space residual for each candidate configuration.

The refinement stage proved to be the most critical component of the entire scheme. Without it, the diffusion model produces structured candidates, but the success rate under strict thresholds of 1 millimeter and 1 degree remains limited. After 50 gradient-refinement steps, the success rate increases to 97.20 percent. After 200 steps, it reaches 99.40 percent. After 500 steps, all evaluated target poses have at least one generated configuration satisfying both thresholds. Crucially, the mean joint-space variability remains 9.77 radians throughout refinement, indicating that the method does not collapse the set of candidates into a single configuration but locally corrects each sample toward the solution manifold while preserving diversity.

In a unified benchmark against damped least squares, KDL, direct regression, and a normalizing-flow baseline, DiffusionIK with refinement achieved 99.80 percent success on 500 unseen poses with 95 percent confidence intervals of 98.9 to 100.0. The same refinement applied to the regression and normalizing-flow baselines yields only 97.40 and 13.80 percent respectively, confirming that the generative stage, and not refinement alone, is decisive. The normalizing-flow baseline, trained with 4.4 million parameters for 300,000 steps on the same procedurally generated data, produced candidates with a median position error of 157 millimeters even after refinement, because candidates starting in the wrong basin of attraction of the task-space residual cannot be recovered by local gradient descent.

Robustness near kinematic singularities was evaluated on a second set of 500 poses selected for having the lowest Yoshikawa manipulability index from a 40-fold pool of reachable poses. Near singularities, the iterative numerical solvers pay either time or reliability: damped least squares retains 99.00 percent success but its median query time grows to 243 milliseconds, while the KDL solver keeps sub-millisecond latency but drops to 91.60 percent. DiffusionIK with refinement retains 98.40 percent success at an essentially unchanged 19 milliseconds per pose, with a median position error of 1.7 millimeters. The fraction of individually converged candidates decreases from 63.5 to 32.3 percent, indicating that ill-conditioning slows individual refinements, while the redundancy of 50 candidates per pose preserves the per-pose success rate.

The computational analysis shows that most of the time is spent on gradient refinement. Denoising Diffusion Implicit Models sampling with 50 steps forms 50 candidates in approximately 100 milliseconds, while refinement with 200 steps requires approximately 1,200 milliseconds on an NVIDIA RTX 4090. The complete generation and refinement scheme takes approximately 1,300 milliseconds for one target pose. The method can operate in several regimes: generative sampling alone provides structured candidates in about 100 milliseconds for initialization of subsequent motion planning, reduced refinement with 50 steps achieves 97.20 percent success in about 300 milliseconds, and full refinement with 200 steps provides 99.40 percent success in about 1,300 milliseconds.

The practical value of the proposed method lies in obtaining not one but a set of diverse kinematically admissible solutions for a single target pose. Such configurations can be used for subsequent selection of a manipulator posture with regard to joint limits, singular configurations, self-collisions, obstacles in the workspace, and requirements imposed by the next motion. The team notes that the current implementation does not include collision checking with the environment or self-collision constraints, and validation on a physical manipulator is needed to account for calibration errors, sensing uncertainty, and actuator-level deviations. Further research should focus on integrating collision constraints into the generation and refinement process, reducing the number of refinement steps, and applying progressive distillation to accelerate diffusion sampling.

Subject of Research: Conditional diffusion-based inverse kinematics for redundant 7-DOF manipulators with differentiable refinement

Article Title: Conditional diffusion-based inverse kinematics for redundant 7-DOF manipulators with differentiable refinement

Article References: Hentosh, L., Hladun, Y., & Tsmots, I. (2026). Conditional diffusion-based inverse kinematics for redundant 7-DOF manipulators with differentiable refinement. Frontiers in Artificial Intelligence, 9, Article 1884749. https://doi.org/10.3389/frai.2026.1884749

Image Credits: AI Generated

DOI: 10.3389/frai.2026.1884749

Keywords: Conditional, diffusion-based, inverse, kinematics, redundant, manipulators, differentiable, refinement, scientific research

Motion deficits, whether caused by stroke, spinal cord injury, diabetes, or neurological disease, have severe impacts on quality of life, and the technologies designed to minimize them sit at the intersection of engineering, neuroscience, and the social sciences. Rehabilitation is, at its core, a re-training process: a feedback loop that depends on observing movement and acting on it. That makes sensors, mechanisms, actuators, and control systems the essential ingredients, and it explains why virtually any modern technology can find a role in the field. Sequeira’s review frames this sprawling domain through taxonomies that classify devices by function, from robotic wheelchairs and smart walkers to wearable exoskeletons, and by the level of automation, the type of human interaction, and the technology readiness level of each system.

The taxonomy question is more than academic housekeeping. Structured classification schemes give clinicians and designers a common language for matching devices to patient needs, and the literature has proposed many of them, from walking support systems for the elderly to broader mobility assistance devices. For physical rehabilitation, the simplest schemes divide the field into exoskeletons, which wrap around the body, and end-effector devices, which interact with a limb at a single point. For mental and neurological rehabilitation, researchers have recognized the need for frameworks that link theoretical models of the brain to clinical practice, connecting stability, muscle performance, and neuromuscular coordination to the sensors and actuators that can measure and assist them.

One of the most striking technical insights in the review concerns kinematics, the mathematics of movement. The adult human body contains 206 bones and, from a robotics perspective, an enormous number of degrees of freedom, the independent directions in which joints can move. Modeling a full walking gait has required as many as 36 degrees of freedom, while studies of the lower limbs alone have used 18. Yet rehabilitation devices exploit the body’s kinematic redundancy to work with dramatically simpler mechanics. Typical lower-limb exoskeletons range from six to fourteen degrees of freedom, upper-limb systems from two to six, and a knee reeducation mechanism can be reduced to just five. The Robear nursing robot, designed to lift patients between a bed and a wheelchair, uses two human-like arms with force sensors on a mobile base, totaling fifteen degrees of freedom. Simpler kinematics, the review emphasizes, brings natural advantages in sensing, actuation, and control, and it is the key to keeping costs down.

Nowhere is that philosophy clearer than in the review’s first case study: a robotic cane. Building on a lineage of robotized canes that stretches from Lidar-equipped, three-wheeled designs for visually impaired users to a lightweight 1.5-kilogram model with an RGB-D camera and ultra-wideband positioning tags, Sequeira’s team built a device with a single degree of freedom, essentially an inverted pendulum mounted on a wheel. When a person walks with a traditional cane, the system alternates between two states: the cane swinging like a pendulum and the cane planted on the ground as a fixed support. A robot cane eliminates that switching entirely, because the wheel never leaves the ground. With only one control variable, the rotation of the wheel, the designer must trade off orientation error against linear displacement, adjusting the cane’s tilt and speed to match the user’s gait. A later version with a deformable wheel smoothed movement over slightly irregular surfaces without compromising stability, and rehabilitation sessions were conducted at two branches of the Portuguese healthcare system with informed, anonymized participants.

The human response was measured with a deliberately minimal questionnaire, a three-point Likert item plus an open question, designed to avoid survey fatigue among elderly participants. Half of the respondents reported a positive attitude toward the robot cane, a quarter were neutral, and a quarter negative. That modest but real acceptance, achieved outside the laboratory, is precisely what the review identifies as a distinctive feature: many published robotic canes have never been tested with actual people with locomotion limitations, and their bulk can make them impractical in cluttered domestic settings where rehabilitation actually happens.

The second case study attacks one of diabetes most devastating complications. Diabetic Foot, characterized by poor circulation in the feet and frequent ulcerations that can lead to amputation, affects an estimated 18.6 million people each year, against a backdrop of roughly 3.4 million diabetes-related deaths in 2024. Standard monitoring relies on visual inspection, sensitivity tests, pedal pulse detection, and pressure analysis, all dependent on the experience of the technician. The review describes an automated alternative developed in the DFAA project at a Portuguese hospital: four images of each patient’s feet are fed to a ResNet50 convolutional neural network trained on a dataset of 3,255 images labeled with the help of generative AI, achieving 80.5 percent agreement with human experts across 87 participants. The system produced F1-scores of 76.51 percent, 72.08 percent, and 89.74 percent for low, medium, and high risk respectively, and a Grad-CAM activation map shows clinicians exactly which regions of the image drove each prediction, a basic but visually intuitive form of explainability. Thermography adds another low-cost signal, since people with peripheral arterial disease tend to have cooler lower limbs, and AI analysis of thermal images has reported accuracies between 85 and 100 percent for detecting ulcers in prior studies.

The third demonstration turns to the brain. Electroencephalograms, the squiggly traces recorded from the scalp, are notoriously difficult to read, and neurologists’ visual analysis is strongly shaped by experience. Research has shown that even normal-looking EEGs can hide epileptic traits such as high-frequency oscillations, spikes, and altered brain connectivity, with machine learning approaches to biomarker detection reporting accuracies between 86 and 99 percent. Sequeira’s team took a radically simple approach: a three-layer convolutional network fed with tiny 28-by-28-pixel images sampled directly from the EEG curves, mimicking the way human readers scan the traces across frequency bands for sharp peaks and asymmetries. Trained on 35 abnormal and 85 normal EEGs, the system achieved a 77 percent median accuracy across 909 runs, with precision of 0.9 and recall of 0.7, and a low interquartile range indicating consistent performance. For a recommendation system that flags recordings for expert review, the authors argue, that is an interesting result, and the misclassified samples illustrate a genuine clinical truth: epileptic traits can hide in apparently normal recordings.

Perhaps the most sobering part of the review is its discussion of why such technologies struggle to reach patients. Performance measured in controlled laboratory settings routinely overstates real-world effectiveness, where clinicians may record only a subset of required data, images may be taken from uninformative angles or under poor lighting, and no one is available to supervise acquisition. Poor data quality, the paper argues, should be treated as a genuine characteristic of the real world rather than as noise. The novelty factor compounds the problem: acceptance of a new device is often high at first and then declines, biasing short experiments. Clinicians have reported low acceptance of rehabilitation robots despite good research indicators, citing safety perceptions, training, and maintenance concerns, while physical therapists have described missed expectations with wearable knee rehabilitation technology. Even pricing psychology works against cheap devices, since low prices can induce a perception of low quality, and consumers tend to weight negative properties more heavily than positive ones. Socio-cultural context matters too, with acceptance differing across studies conducted in different environments.

The review’s conclusions are pragmatic rather than utopian. Despite the staggering kinematic complexity of the human body, very simple mechanics can yield genuinely useful rehabilitation systems, and AI techniques have a clear role in helping technicians analyze data and cope with the variability of human bodies and health conditions. The real challenge, Sequeira argues, lies in developing agile experimental protocols that distinguish early medical viability studies from certification experiments, so that innovation can diffuse through healthcare systems as fast as it does through consumer technology. In ageing societies and struggling economies alike, the devices that matter most may not be the most sophisticated ones, but the cheapest ones that actually work in the messy, uncontrolled environments where people live and recover.

Subject of Research: Low-cost robotics and artificial intelligence for rehabilitation and health monitoring

Article Title: Low-cost robotics and AI technologies for rehabilitation and health monitoring

Article References: Sequeira, J. S. (2026). Low-cost robotics and AI technologies for rehabilitation and health monitoring. Frontiers in Robotics and AI, 13, Article 1919797. https://doi.org/10.3389/frobt.2026.1919797

Image Credits: AI Generated

DOI: 10.3389/frobt.2026.1919797

Keywords: rehabilitation robotics, low-cost devices, artificial intelligence, robotic cane, diabetic foot, EEG, machine learning, exoskeletons, wearable sensors, health monitoring, human factors, kinematics

The platform, called OpenCap, fuses synchronized smartphone videos with musculoskeletal modeling to estimate three-dimensional human movement without any of the reflective markers that traditional systems require. Conventional three-dimensional motion capture, the long-standing reference standard, demands an array of specialized cameras, dozens of markers glued to the skin, trained operators, and dedicated space. Those requirements have kept sophisticated gait and movement analysis confined to research laboratories, even as clinicians increasingly recognize that a single completion time for the five-times sit-to-stand (5STS) test conceals important differences between patients who may arrive at the same total time through very different movement strategies.

Led by Jialiang Xu and Qitao Tan, with Ming Zhang as senior collaborator, the research team recruited twenty-four older Asian women aged 65 and older. Half were classified as healthy and half as sarcopenic, meaning they had the age-associated loss of skeletal muscle mass and function that affects an estimated 10 to 40 percent of older populations and contributes to frailty, falls, and disability. Classification followed the 2025 consensus criteria of the Asian Working Group for Sarcopenia, requiring both low handgrip strength measured with a handheld dynamometer and low height-adjusted appendicular muscle mass measured by bioelectrical impedance. The cohort deliberately focused on women to maintain homogeneity within this sex-specific classification framework, a constraint that shapes how broadly the findings can be generalized.

Each participant performed the 5STS task on a standardized 43-centimeter armless chair with arms crossed over the chest, completing four recorded trials of five consecutive sit-stand-sit repetitions. While an eight-camera Vicon system tracked thirty-nine reflective markers at 250 hertz, three iPhone 14 devices mounted on tripods recorded 720-pixel video at 60 frames per second from frontal, oblique, and lateral viewpoints. The OpenCap pipeline processed the videos using the HRNet pose-estimation algorithm, generating subject-specific musculoskeletal models and extracting hip, knee, and ankle flexion-extension angles. To ensure a fair comparison, the researchers fed both systems through the same anatomical model geometry and the same 6 hertz low-pass filter before analyzing the waveforms.

The agreement results were striking. For knee flexion and extension, OpenCap tracked the laboratory system with a root mean square error of just 1.84 degrees and an intraclass correlation coefficient of 0.998, a level of near-perfect absolute agreement. Hip flexion and extension also showed high agreement, at 0.970, though with larger absolute errors of 5.42 degrees. The ankle proved the weak link: despite small angular errors, its agreement was more variable, which the authors attribute partly to the limited ankle excursion the task demands. Notably, agreement was calculated from participant-level mean waveforms, and the authors caution that these estimates should not be extrapolated to single trials or repetitions.

The validation came with important technical caveats that the authors openly discuss. Because the two recording systems lacked a shared hardware trigger, they were aligned using cross-correlation of the knee waveform, which happened to be the same joint later used to judge agreement, potentially favoring the knee result. Eight of the ninety-six recorded trials were also excluded due to OpenCap synchronization failures that made pose tracking impossible, a reminder that real-world failure rates matter for any technology proposed for clinical screening. Still, the overall message was that for the two joints that dominate the chair-stand movement, the smartphone system reproduced laboratory kinematics with impressive fidelity.

The second half of the study asked whether this rich kinematic data could actually distinguish sarcopenic from healthy women beyond what a stopwatch shows. Using statistical parametric mapping, a technique that tests for differences across entire movement trajectories while controlling the family-wise error rate, the researchers found that the sarcopenic group showed reduced hip flexion near both the beginning and end of the sit-stand-sit cycle, along with brief localized reductions in knee flexion. Angular velocity waveforms, however, did not differ significantly between groups, possibly because the longer movement durations in the sarcopenic group spread the same angular displacement over more time.

The temporal measures told an even clearer story. Sarcopenic participants took significantly longer to complete the full five stands, and the analysis of individual movement phases revealed that both the ascent, the effortful rise requiring lower-limb extensor force generation, and the descent, which demands eccentric control and postural stability, were prolonged. Perhaps most intriguingly, the sarcopenic group showed greater progressive slowing across the five repetitions, quantified as a steeper repetition-duration slope, while their repetition-to-repetition variability was no different from the healthy group. That combination suggests a systematic decline in pace, consistent with reduced functional reserve or fatigability, rather than mere temporal irregularity, though the authors emphasize that mechanisms such as reduced eccentric strength or fear of falling remain untested hypotheses in this study.

The clinical implications are tantalizing but carefully bounded. The authors note that individuals with similar total completion times could differ substantially in rising duration, lowering duration, progressive slowing, or joint-movement strategy, and phase- and repetition-specific analysis could pinpoint where functional limitations actually occur within the task. Yet this study did not test whether these derived measures improve diagnosis, prognosis, or treatment selection beyond conventional timing, and the OpenCap-derived temporal measures were not validated against an independent external reference. Before routine use, researchers will need evidence on repeatability, clinically meaningful thresholds, processing time, and acceptability in community and primary-care settings.

The study’s limitations also frame its scope. The sample was small, restricted to older Asian women with a relatively lean anthropometric profile, and the between-group analyses were unadjusted for potential confounders such as physical activity, medication use, and pain. The cross-sectional design cannot support causal inference, and no direct measures of muscle activation, joint kinetics, or physiological cost were collected. Even so, the work marks a meaningful step toward democratizing biomechanical assessment. If three smartphones and open-source software can capture how aging muscles struggle with something as fundamental as rising from a chair, the door opens to large-scale screening in communities where reflective markers and eight-camera laboratories will never reach.

Subject of Research: Smartphone-based markerless motion capture validation and sarcopenia assessment using the five-times sit-to-stand test in older Asian women

Article Title: OpenCap-based five-times sit-to-stand analysis in older Asian women with and without sarcopenia

Article References: Xu, J., Cai, H., Wang, Y., Hu, Z., Zhang, M., & Tan, Q. (2026). OpenCap-based five-times sit-to-stand analysis in older Asian women with and without sarcopenia. Frontiers in Aging, 7, Article 1960201. https://doi.org/10.3389/fragi.2026.1960201

Image Credits: AI Generated

DOI: 10.3389/fragi.2026.1960201

Keywords: OpenCap, sarcopenia, five-times sit-to-stand, markerless motion capture, biomechanics, older adults, kinematics, smartphone cameras, musculoskeletal modeling, statistical parametric mapping, AWGS 2025, aging

Johannes Keck, Kazuyuki Sato and Rouwen Cañal-Bruland of Friedrich-Schiller-University Jena set out to close a persistent gap in the science of action understanding. Most previous experiments examined movement cues and contextual inference in isolation, often asking passive observers to make explicit judgments about pre-recorded actions. Real social life is nothing like that. Interacting partners respond to each other continuously, in real time, while the information available to them shifts from moment to moment. To capture this dynamism, the team turned to an unlikely but elegant experimental model: the children’s game rock-paper-scissors. As a zero-sum game with a cyclic, non-transitive payoff structure—rock beats scissors, scissors beats paper, paper beats rock—it combines genuine competition, multiple action alternatives and a rich history of play, all of which unfold within roughly a second and a half per game.

Ninety participants, paired into forty-five dyads, played hundreds of iterative games of rock-paper-scissors while wearing full-body motion capture suits studded with fifty-seven reflective markers. A sixteen-camera near-infrared system recorded their movements at 240 hertz, allowing the researchers to reconstruct the characteristic up-and-down wrist swings in exquisite detail. Each game was segmented into three phases corresponding to the first, second and final swing of the wrist, and eleven movement features—ranging from peak velocity and acceleration to trajectory skewness, positional variability and a measure of velocity unpredictability called velocity entropy—were extracted for every phase. After preprocessing, thirty-eight dyads with complete, reliably reconstructed trajectories entered the final analysis.

The first question was whether intention information is actually embedded in these naturally generated movements. To find out, the researchers trained a multinomial logistic classifier to map the phase-specific kinematic features onto the action each player ultimately performed. The classifier decoded the intended gesture with an accuracy of 47.6 percent, significantly above the 33.3 percent chance level. Crucially, this encoding accuracy did not differ between winning and losing trials, nor across the three movement phases. In other words, every player’s wrist leaked information about their upcoming choice, whether that choice would ultimately triumph or fail. The information was there for the taking, in every game, at every stage of the unfolding movement.

The surprise came when the researchers asked whether opponents actually exploited this information. Because explicit perceptual reports would have disrupted continuous play, the team derived an indirect, choice-based measure they call likely kinematic readout: if a player chose rock, for example, the analysis inferred that the player expected the opponent to throw scissors, and then tested whether the opponent’s kinematics on that trial actually supported that inference. Overall, likely readout exceeded chance, reaching 39.2 percent. But the outcome-dependent pattern was striking. On losing trials, readout was significantly above chance in every phase and rose steadily as the game progressed, peaking at 38.6 percent in the final swing. On winning trials, it never exceeded chance at all and actually declined to 31.1 percent by the end. Players who consistently followed the movement-derived signal across all three phases lost significantly more often than those who did not.

At first glance this seems paradoxical: why would reading the opponent’s body better make you lose? The authors argue that in fast-paced competitive settings with multiple possible outcomes, kinematic cues are noisier and more ambiguous than in the controlled, cooperative tasks used in most prior studies. Opponents may also deliberately disguise their intentions, converting kinematic information into kinematic disinformation, as research on deceptive movements in sports has repeatedly shown. Under such conditions, generating several competing kinematic predictions can reduce reliability despite high perceptual sensitivity, and observers may do better to complement—or override—movement evidence with contextual cues drawn from the game’s structure and history.

To test this idea formally, the researchers modeled each player’s contextual inference using four competing computational models drawn from two conceptual families. Action-based models predicted choices from sequential regularities in previous actions: a first-order Markov model tracking transition probabilities between consecutive gestures, and the cognitive hierarchy assessment model, or CHASE, which formalizes recursive strategic reasoning across levels of sophistication. Outcome-based models adapted predictions according to wins and losses: a win-stay/lose-shift model and a reinforcement-learning model implementing a Rescorla-Wagner update rule with softmax action selection. These model-derived posteriors were then fed, together with the kinematic measures, into hierarchical mixed-effects logistic regressions predicting single-trial winning.

The results were unambiguous. Reinforcement learning was the strongest positive predictor of winning: trials in which the model assigned high posterior probability to the player’s chosen action were significantly more likely to end in victory, with an odds ratio of 1.37. Win-stay/lose-shift contributed a smaller positive effect, whereas reliance on first-order Markov sequential regularities actually predicted reduced winning, suggesting that players who exploit simple action patterns become more exploitable themselves. On the kinematic side, richer intention information encoded in one’s own movement was modestly beneficial, but stronger behavioral alignment with the likely kinematic readout significantly reduced the odds of winning. Both information sources contributed unique predictive value, but context explained substantially more trial-to-trial variation in success than movement did.

The most compelling evidence came from conflict trials—decisive games in which the contextual model and the likely kinematic readout recommended different actions. Here, success can be attributed to one source or the other, and context won decisively. Following the reinforcement-learning recommendation more than doubled the odds of winning relative to following the kinematic signal, with an odds ratio of 2.33. Descriptively, participants won 60.6 percent of disagreement trials when they followed the reinforcement-learning recommendation, compared with only 42.7 percent when they followed the kinematic readout—a 17.9 percentage point advantage that significantly exceeded the benefits of the Markov, win-stay/lose-shift and CHASE models. An interaction analysis reinforced the picture: the behavioral influence of kinematic readout became increasingly detrimental as reinforcement-learning evidence accumulated, while the contribution of kinematic encoding itself remained additive across contextual states.

Taken together, the findings suggest that successful action understanding in dynamic social interaction is not a matter of maximizing sensitivity to bodily signals but of adaptively arbitrating between information sources. Movement kinematics consistently carried intention-related information, yet its behavioral value depended on the contextual state in which it was interpreted. The study also refines earlier neural work suggesting that tracking recent outcomes can impair performance in rock-paper-scissors: rather than contextual information being generally maladaptive, different forms of context have markedly different consequences, with reward-based learning outperforming simple sequential dependencies. The authors caution that their readout measure is an indirect behavioral proxy, and they propose that immersive virtual reality with systematically manipulated avatar kinematics could validate it more directly. For now, the message is clear: in the heat of real-time competition, the body talks, but the smartest listeners weigh what it says against everything else they know.

Subject of Research: Kinematic and contextual information integration in real-time competitive social interaction

Article Title: Context shapes action understanding in social interactions

Article References: Keck, J., Sato, K., & Cañal-Bruland, R. (2026). Context shapes action understanding in social interactions. iScience, 29(11), Article 117796. https://doi.org/10.1016/j.isci.2026.117796

Image Credits: AI Generated

DOI: 10.1016/j.isci.2026.117796

Keywords: action understanding, kinematics, social interaction, rock-paper-scissors, reinforcement learning, motion capture, action prediction, deception, computational modeling, decision making, contextual inference, iScience

The core problem the team set out to solve is a familiar one in sports robotics. Exoskeleton robots, the most common type of wearable training aid, attach rigid links and motors directly to the athlete’s joints. That adds substantial inertia to the limbs, which is precisely what a speed skater does not need. A bulky frame strapped to the hips and legs can constrain the very fluid, high-speed movements the athlete is trying to refine. Cable-driven robots offer an elegant alternative: the motors sit on a stationary chassis, and lightweight, high-strength cables transmit forces to the athlete’s waist, legs, and ankles. The cables decouple the heavy drive hardware from the human body, and their inherent compliance softens the risk of rigid collisions. The trade-off is that controlling eight cables to move a platform in six degrees of freedom is a genuinely hard control problem, and existing control schemes were not precise enough for competitive skating.

The robot itself is built around a gantry of vertical columns and crossbeams, with pulleys guiding the cables along smooth force-transmission paths. Eight functional modules make up the machine: the load-bearing columns, the crossbeams that stabilize the gantry, the pulleys, ergonomically curved handrails that give the skater balance support, servo drivers housed in the chassis with high-precision encoders providing millisecond-level feedback, lightweight composite cover plates for protection and quick maintenance, a chassis integrating counterweights and shock absorption, and a polished mirror surface whose friction coefficient is tuned to resemble real ice. Cables connect to key points on the athlete’s waist, legs, and ankles, forming a closed-loop force control network. The design accommodates athletes from 1.65 to 1.98 meters tall, with a drive unit capable of 380 newtons of pulling force and a suspension rated for 500 kilograms, margins engineered to withstand the explosive peak loads of a skater’s push-off.

The drive unit combines an S7-1200 programmable logic controller, a 400-watt permanent magnet synchronous motor rated at 3000 revolutions per minute, and a type 1204 ball screw. When the motor spins, it turns the screw, which drives a slider that reels a cable in or out, adjusting both its length and tension in real time. This hybrid scheme lets the robot provide upward lifting assistance during the take-off phase and apply controllable damping at the moment of ice landing, cushioning joint impact. Because the motors live on the chassis rather than on the athlete, the components attached to the body stay light, preserving the flexibility that rigid exoskeletons sacrifice.

To make the machine useful, the researchers first had to solve its kinematics: the mathematical mapping between the pose of the moving platform, which carries the skater’s leg brace, and the lengths of the eight cables. They established a fixed coordinate system at the geometric center of the robot’s base frame and a moving coordinate system fixed to the platform at the center of the athlete’s thigh. Assuming the cables remain fully tensioned and can be treated as ideal massless straight segments, they derived the inverse kinematics, which computes cable lengths from a desired pose, using a closed vector quadrilateral approach combined with a rotation matrix built from roll, pitch, and yaw angles. Forward kinematics, the harder reverse problem of recovering the platform’s position and orientation from known cable lengths, was solved with the Newton-Raphson iterative method, chosen over analytical approaches for its rapid convergence and precision.

The Newton-Raphson method converges quadratically near the true root, but it is sensitive to the initial guess. The team exploited a practical trick: because the servo control system samples at millisecond intervals, the platform barely moves between consecutive control cycles, so the pose from the previous cycle serves as an excellent starting point for the next iteration. This guarantees robust, real-time convergence. To validate the model, the researchers ran simulations in MATLAB using five sets of pose points within the robot’s workspace, feeding them through inverse kinematics to get cable lengths and then through forward kinematics to recover the poses. The maximum deviation between the original and recovered poses was 0.95 percent, below the 1 percent threshold the team set as evidence that the kinematic model is sound.

With the mechanics settled, the heart of the paper is its control strategy. Conventional proportional-integral-derivative controllers, while simple and robust, struggle with the nonlinear coupling and sudden disturbances that characterize skating, such as abrupt changes in ice friction or the athlete’s own movements. Active disturbance rejection control offers a way forward: an extended state observer continuously estimates the combined internal and external disturbances acting on the system and compensates for them in real time. But traditional ADRC architectures include a tracking differentiator, a component that shapes the response to step inputs to avoid overshoot by ramping the control output up gradually. That conservatism introduces delays and sluggish response, a poor fit for the high-frequency movements of skating, like rapid ice pushes and sudden stops.

The researchers’ solution is a fractional-order active disturbance rejection controller, or FOADRC, with two key innovations. First, they removed the tracking differentiator entirely, allowing the large initial error to drive the system promptly and sharpen the dynamic response. Second, they replaced the conventional state feedback error law with a fractional-order PD control law and upgraded the observer to a fractional-order extended state observer. Fractional-order calculus, which uses non-integer differentiation and integration operators, gives the controller adjustable amplitude-frequency slopes and better high-frequency noise suppression than integer-order designs, along with superior robustness to complex disturbances. The fractional-order observer was tuned using a bandwidth parameterization borrowed from integer-order observer theory, so that only a single parameter needs adjustment, dramatically simplifying calibration. The fractional-order operators themselves were approximated in a practical frequency band using the Oustaloup method, a standard rational-fitting technique accurate enough at fifth order.

To generate realistic control targets, the team captured real skating motion using a NOKOV infrared motion capture system with eight cameras sampling at 100 hertz and a spatial positioning error under 0.1 millimeters. A 25-year-old male speed skater, 178 centimeters tall and weighing 70 kilograms, performed standard straight-line skating cycles while retroreflective markers on his hip, thigh, lower leg, and ankle tracked the movement. The researchers focused on the hip joint, the hub connecting trunk and legs that carries body weight and fine-tunes the center of gravity during skating. Raw marker data were smoothed with a fourth-order zero-phase Butterworth low-pass filter at a 6 hertz cutoff and gaps from marker occlusion were filled with cubic spline interpolation. The resulting hip joint angle time series, covering sagittal-plane flexion-extension and coronal-plane abduction-adduction, was fitted with an eighth-order Fourier series to produce a smooth, reproducible reference trajectory, and measured cable tensions grounded the simulation in realistic loads.

The payoff came in comparative simulations against a traditional PID controller, evaluated by motor angle tracking error on the drive units for the first two cables. The FOADRC strategy narrowed the error range substantially and improved motion control precision, and, crucially, the tracking error profiles of the two motors were highly consistent despite following different trajectories and load variations. That uniformity matters enormously in multi-cable systems, where all drives must stay synchronized to avoid unbalanced internal tensions or uncoordinated movements. The authors are candid about the study’s limits: everything so far rests on simulation and data from a single subject. The next step is building a physical prototype and running multi-subject trials in real training environments. If those succeed, the combination of flexible cable actuation and fractional-order disturbance rejection could become a template not just for skating, but for a broader generation of intelligent training equipment across competitive sports.

Subject of Research: Cable-driven skating training robot kinematics and fractional-order active disturbance rejection control

Article Title: Research on kinematics and fractional order active disturbance rejection control of skating training robot

Article References: Wang, B., Zhao, X., Gong, Y., Sun, L., & Yang, Z. (2026). Research on kinematics and fractional order active disturbance rejection control of skating training robot. Mechanical Sciences, 17(2), 731-746. https://doi.org/10.5194/ms-17-731-2026

Image Credits: AI Generated

DOI: 10.5194/ms-17-731-2026

Keywords: speed skating, cable-driven robot, fractional-order control, active disturbance rejection, kinematics, Newton-Raphson method, permanent magnet synchronous motor, motion capture, sports robotics, trajectory tracking, training technology, Mechanical Sciences

The core idea is deceptively simple. Where the classic cube relies on intricate spatial axis layouts and intersecting revolute joints, the PPRCM replaces rotation with pure translation. Its rigid components slide along grooves machined into a flat base, arranged in a grid reminiscent of the sliding-tile puzzles that have entertained generations. The researchers point out that this planar constraint is a genuine engineering advantage: planar mechanisms are far easier to actuate, analyze, and manufacture than their spatial counterparts, because theoretical derivation and numerical calculation are dramatically simplified when all motion happens in two dimensions.

The team built their theory around a physical prototype, also rendered as a three-dimensional printed model, consisting of a base and three distinct moving modules. The base carries uniformly distributed grooves along two perpendicular directions, allowing the modules to translate along either axis. The design borrows from an unexpectedly ancient source: the tenon-and-mortise joints of traditional Chinese woodworking, which inspired the groove-and-pin interlocking scheme that dynamically connects modules to the base. Six moving paths run along each direction, giving the mechanism a rich lattice of possible trajectories.

Each moving module is itself a layered assembly. It contains a moving block together with moving-block side pieces, or MBSPs, fitted with clamping pins that keep the assembly engaged in the base grooves during translation. The side pieces also carry their own internal grooves, which lock the moving block onto them when it slides along the horizontal direction. This creates three distinct motion modes: a module can translate along one axis, along the perpendicular axis, or the inner block can slide along the horizontal direction on top of its own side pieces. Because the contact surfaces differ between the first two modes and the third, the third mode offers the richest design possibilities, and the researchers chose it as the focus of their theoretical analysis.

That analysis centers on a concept the authors call the prismatic-pair contact surface, or PPCS. A prismatic pair is the fundamental sliding joint of mechanism theory, permitting exactly one translational degree of freedom between two components. The PPCS is the direct contact area between those components, and because it is generated by extruding a two-dimensional profile along the direction of motion, it can be uniquely described by a single profile equation in a cross-sectional coordinate system. The researchers classify these surfaces as symmetric or asymmetric depending on whether a plane of symmetry parallel to the motion exists, and they restrict their formal treatment to the symmetric case for clarity.

From this foundation, the team derived the conditions under which components can actually swap positions and reconfigure the mechanism. Three rules emerge. First, components must share identical movement slopes, meaning only parts translating along the same axis are mutually interchangeable. Second, the contact surfaces must satisfy an inclusivity requirement: a moving component can slide on a fixed component only if the fixed component’s profile contains the mover’s profile, a relationship the authors express mathematically and prove to be transitive. Third, the mechanism must contain at least two intersecting movement directions, because with only one direction all motion is parallel and no component can ever move in two ways, which would make reconfiguration impossible.

These conditions feed directly into the paper’s central contribution: a type synthesis method for the PPRCM. Type synthesis is the branch of mechanism theory concerned with enumerating the structural forms a machine can take, independent of its exact dimensions. The method proceeds in four steps: choose the number of movement directions, choose how many parallel moving paths exist in each direction, choose the types of contact surfaces in each direction, and finally design a base on which all movable components can be placed. Because these three factors, direction count, path count, and surface type, fully determine the mechanism’s topology, specifying their values generates a complete family of distinct PPRCM designs.

To demonstrate the method’s power, the authors worked through systematic case studies. They generated configurations with two, three, and even four different movement slopes, including versions whose directions are not perpendicular but angled at 45 degrees, showing that the theory accommodates non-orthogonal grids. They varied the number of moving paths per direction from two up to six, and they swapped in different sets of contact-surface profiles with different inclusivity hierarchies. For every synthesized configuration, corresponding three-dimensional models were constructed to validate that the designs are geometrically rational and mechanically effective, confirming that the method reliably produces working reconfigurable mechanisms rather than paper abstractions.

The significance extends well beyond puzzle-inspired novelty. Rubik’s cube mechanisms have already been combined with origami structures, deployable space hardware, and modular robots, and cube-inspired mathematics underpins image encryption schemes, biometric template protection, microfluidic devices, and even deep-learning architectures that exploit the cube’s group structure. A planar version with a rigorous synthesis theory gives engineers a tractable starting point: it simplifies the cognitive and analytical difficulty of complex reconfigurable mechanisms and lays a foundation for extending the approach to fully three-dimensional spatial designs. Because reconfigurable machinery is central to deformable structures, adaptable robots, and variable working scenarios, a systematic recipe for generating new configurations is a valuable tool.

The authors also situate their work in a long intellectual lineage. Planar type synthesis has evolved from designers’ intuition into formal methods based on kinematic chain enumeration, finite group theory, Assur groups, matrix representations, and, more recently, neural-network and big-data approaches. Yet none of these established methods transfers directly to Rubik’s cube mechanisms, whose unique sliding, interlocking, and position-swapping behavior demands new theory. The PPRCM framework also echoes much older ideas about rearrangement: the authors trace the concept back to Luo Shu, the ancient Chinese symbolic numerical pattern embodying transformation and rearrangement, and to sliding puzzles such as the eight-digit game and the fifteen-puzzle, which are, in mechanism terms, planar prismatic systems. By formalizing what those puzzles do intuitively, the study enriches planar mechanism theory and opens a path toward a new generation of flat, reconfigurable machines whose configurations can be counted, predicted, and designed on demand.

Subject of Research: Reconfigurable kinematic analysis and type synthesis of planar prismatic Rubik's cube mechanisms

Article Title: Research on reconfigurable kinematic analysis and type synthesis of planar prismatic Rubik's cube mechanism

Article References: Fan, D., Zeng, D., Zhao, Y., Feng, H., Deng, Y., & Liu, Y. (2026). Research on reconfigurable kinematic analysis and type synthesis of planar prismatic Rubik's cube mechanism. Mechanical Sciences, 17(2), 813-823. https://doi.org/10.5194/ms-17-813-2026

Image Credits: AI Generated

DOI: 10.5194/ms-17-813-2026

Keywords: reconfigurable mechanisms, Rubik's cube mechanism, type synthesis, planar mechanism, prismatic pair, kinematics, topology, mechanism design, sliding puzzle, metamorphic mechanism, mechanical sciences, 3D printed prototype

In a preprint posted on 29 September 2026 and currently under review for the journal Mechanical Sciences, Dabao Fan and colleagues at Suqian University and Dongguan University of Technology describe an adaptive grasping device whose fingers are built from improved parallelogram units. The key innovation lies in how the classic mechanism is modified. In a standard parallelogram linkage, opposite links stay parallel throughout motion, which is precisely why the geometry is so useful for transmitting predictable motion. The researchers took this familiar arrangement and introduced elastic elements, allowing the linkage to deform passively under load. When the gripper closes on an object, the springs within each parallelogram unit yield in a controlled way, letting the finger surfaces conform to the object’s size and shape without any additional sensing or control effort.

This passive adaptability is the conceptual heart of the design. Rather than measuring an object and computing a grasp, the device lets physics do the computation. The elastic deformation of the springs acts as a mechanical buffer: stiff enough to hold a workpiece securely, compliant enough to absorb dimensional variation and protect fragile or flexible items from excessive contact forces. According to the authors, this allows the gripper to effectively adapt to objects of different sizes while preserving the integrity of the grasped workpieces, addressing two of the most persistent complaints about traditional rigid grippers: poor environmental adaptability and easy damage to delicate goods.

Adaptive fingers, however, still need to be driven, and this is where the team’s second contribution comes in. Each gripper unit is actuated by a 2-RRR planar parallel mechanism, a configuration in which two chains of three revolute joints each connect a fixed base to a moving platform. Parallel mechanisms of this kind are prized for their stiffness, accuracy, and load capacity, since the load is shared among multiple kinematic chains rather than cantilevered off a single serial arm. The trade-off is usually a complicated analysis problem: understanding exactly how the platform can move requires a careful accounting of the mechanism’s degrees of freedom and its constraints.

The researchers tackled that analysis with screw theory, a mathematical framework that represents both the rotational and translational motions of rigid bodies as screws, combining an axis, a pitch, and an amplitude. Using screw theory together with an explicit kinematic model, the team characterized the degree of freedom and the constraint characteristics of the 2-RRR driving mechanism, establishing rigorously which motions the gripper unit can perform and which are locked out by the parallel architecture. This kind of validation matters in practice, because a gripper whose kinematics are only loosely understood tends to surprise its designers when it meets an object it was not analyzed against.

Perhaps the most immediately practical element of the work is its treatment of actuation. Adaptive grippers with multiple independently moving finger segments often accumulate driving redundancy: many motors, each requiring its own controller, driver electronics, and wiring, all coordinated by software that must keep them synchronized. The authors report that traditional schemes of this family can demand as many as six motors. Their solution is a linkage transmission structure built around bevel gear sets, which mechanically couples the motion so that the entire device is driven in a dual-motor mode. Bevel gears, which transmit rotation between shafts at angles to one another, allow a single input to distribute coordinated motion across the linkage network. The result, the team states, is a significant reduction in driving redundancy and overall control cost, a benefit that compounds over production runs of thousands of units.

To verify that the theory translated into function, the researchers carried out grasping condition analyses for objects of different sizes. These tests confirmed, they report, that the device possesses excellent dimensional adaptability and grasping stability. In other words, the same hardware could close reliably around a range of target dimensions without reconfiguration, and once closed, it held its payload stably, with the spring-loaded parallelogram units maintaining conformal contact rather than fighting against the object’s geometry. For applications such as bin picking, agricultural handling, or assembly of parts with manufacturing tolerances, that combination of range and stability is exactly what determines whether a gripper is usable or a nuisance.

The significance of the work lies less in any single component than in the design philosophy it demonstrates. Soft robotics has made enormous strides by replacing rigid links with elastomers, granular jamming, and pneumatic chambers, but soft hands often sacrifice precision, load capacity, and manufacturability. The Suqian and Dongguan team’s approach stays firmly within the world of rigid links and conventional joints, achieving compliance only where it is needed, through deliberately introduced elasticity. This keeps the device cheap to fabricate, easy to model, and compatible with standard precision manufacturing, while still delivering the gentle, shape-following behavior that soft systems achieve through material compliance. It is a reminder that mechanism design, the oldest branch of robotics, still has room for genuinely new ideas.

The authors position the results as providing a reliable theoretical basis and technical reference for two audiences at once: engineers seeking innovative optimization of traditional mechanical mechanisms, and designers of flexible adaptive grasping equipment for engineering applications. Because the analysis is grounded in screw theory and formal kinematic modeling rather than ad hoc prototyping alone, the framework could be adapted by other groups designing grippers with different link geometries, spring stiffnesses, or actuation layouts. The dual-motor bevel gear transmission, in particular, offers a template for taming the actuation complexity that adaptive mechanisms typically incur.

As the preprint moves through peer review, with its public discussion open until 7 November 2026, the broader robotics community will be watching to see how the design performs beyond the laboratory: under dynamic loads, in three-dimensional grasps where fingers interact with one another, and at production scale. What is already clear is that the humble parallelogram, a linkage familiar to every mechanical engineering student, still has untapped potential when its geometry is paired with the right elasticity. In a field captivated by learning-based control and dexterous five-fingered hands, there is something quietly compelling about a gripper that solves the adaptability problem with springs, gears, and four-bar linkages, no neural network required.

Subject of Research: Adaptive robotic grasping using improved parallelogram mechanisms with passive spring compliance and a 2-RRR parallel drive

Article Title: Design and Analysis of an Adaptive Grasping Device Based on Parallelogram Mechanism

Article References: Fan, D., Feng, H., Zhao, Y., Liu, Y., Deng, Y., & Zeng, D. (2026). Design and Analysis of an Adaptive Grasping Device Based on Parallelogram Mechanism. https://doi.org/10.5194/ms-2026-170

Image Credits: AI Generated

DOI: 10.5194/ms-2026-170

Keywords: robotics, adaptive gripper, parallelogram mechanism, screw theory, kinematics, 2-RRR parallel mechanism, bevel gear transmission, compliant grasping, mechanism design, actuation redundancy, Mechanical Sciences, preprint

The review, led by Joachim D’Hondt and colleagues, was conducted according to the PRISMA 2020 reporting guidelines and registered in advance on PROSPERO. The researchers searched four major databases, PubMed, Scopus, SPORTDiscus and Web of Science, identifying 6,804 records, which reduced to 4,811 unique articles after duplicates were removed. Rigorous title, abstract and full-text screening left just ten studies that met the inclusion criteria: original, peer-reviewed investigations of lower inter-limb asymmetry and sprinting performance in healthy, injury-free sprint athletes. Methodological quality was assessed with a modified Downs and Black Quality Index Tool, and the certainty of the evidence was graded using the GRADE framework. The studies, published between 2013 and 2025, came from Europe, North America, Oceania and Asia, and together covered 684 participants, including sixteen world-class sprinters, eight elite sprinters and 138 highly trained athletes.

Inter-limb asymmetry is not a single phenomenon but a family of them. The review distinguishes functional asymmetries, such as differences in strength or jump performance between legs; morphological asymmetries, such as differences in limb length, joint width or muscle cross-sectional area; kinematic asymmetries, involving discrepancies in joint angles, step length or contact time; and kinetic asymmetries, meaning imbalances in ground reaction forces or impulse production. Each type was measured with different tools across the ten studies, from single-leg countermovement jumps and magnetic resonance imaging to textile electromyography embedded in shorts and force measurements during actual sprinting. Complicating matters further, the studies used at least five different equations to calculate asymmetry, including the symmetry angle, percentage differences between superior and inferior limbs, fluctuating asymmetry indices and composite scores.

The headline finding is stark in its simplicity: in general, only limited associations were observed between any of these asymmetry types and sprinting performance. Prvulović and colleagues found no significant link between explosive strength asymmetry, measured with a single-leg countermovement jump, and mean velocity in a 100-metre sprint. Tottori and colleagues reported no significant associations between asymmetry in the cross-sectional area of the psoas major, quadriceps femoris or hamstrings and curve-sprinting time, with the single exception of psoas major asymmetry, which showed a small negative association with cross-directional sprint time differences. Nagahara and Gleadhill found no significant relationships between kinetic asymmetry metrics such as braking and propulsive impulse and maximal sprinting velocity over 60 metres.

Yet the picture is not uniformly null, and the exceptions are scientifically revealing. Gołaś and colleagues, studying eight elite female sprinters with textile EMG sensors, found weak-to-moderate positive associations between sprinting time and EMG asymmetry in the quadriceps, hamstrings and gluteal muscles, meaning greater asymmetry accompanied slower times. Bissas and colleagues reported that in world-class male sprinters, 100-metre time was strongly negatively associated with foot vertical velocity pre-touchdown asymmetry, while in world-class females, shank angle touchdown asymmetry was positively associated with sprint time, a striking sex-specific pattern. Exell and colleagues even found that net ankle work asymmetry and maximum vertical force asymmetry were positively associated with mean 60-metre velocity, suggesting that in some athletes, asymmetry and speed coexist, or that the dominant limb compensates for the weaker one during propulsive phases.

Perhaps the most famous data point in this debate comes from analyses of Usain Bolt himself. Previous biomechanical investigations showed that the fastest man in history ran with pronounced asymmetries: his right leg produced approximately 13 percent greater peak force, while his left leg recorded ground-contact times roughly 14 percent longer, alongside notable step-length discrepancies. If perfect symmetry were a prerequisite for world-record sprinting, Bolt’s mechanics would have made it impossible. The review’s authors point to computer simulation studies suggesting a plausible mechanism: athletes may adopt compensation strategies in which the dominant limb exerts greater force during propulsive phases to offset the weaker limb, preserving overall output despite the imbalance.

The contrast with team-sport research is equally instructive. Earlier work, including a meta-analysis by Fox and colleagues, reported a small but significant negative association between lower-limb asymmetry and sprinting performance, with correlations around 0.20. But that evidence base drew overwhelmingly from unilateral team sports such as soccer, where eight of nine studies involved players whose repetitive kicking and cutting patterns naturally create larger asymmetries. In one striking example, single-leg countermovement jump asymmetry in English Premier League academy players showed strong positive correlations with 5, 10 and 20-metre sprint times, and asymmetry magnitudes in such cohorts have reached 18 percent. Sprint athletes, by contrast, tend to display much smaller asymmetries, often in the low single digits, which may simply fall below any threshold at which performance would be affected.

The review also delivers a methodological caution that should reshape how future studies are designed. The certainty of evidence across all outcomes was rated as very low, reflecting methodological limitations, imprecision and the observational nature of the studies. Only three of the ten studies reported reliability data for their asymmetry measures, a serious gap given that asymmetry is a ratio metric highly sensitive to measurement error in either limb. The authors also argue against composite asymmetry scores, which aggregate different body regions and metrics into a single number; such scores can conflate large asymmetries in a few variables with small asymmetries across many, obscuring meaningful patterns. Because asymmetry is known to be task-, metric- and segment-specific, assessing each type independently is considered far more valid and informative.

For coaches and athletes, the practical message is one of cautious reassurance. The evidence synthesized here does not support the assertion that inter-limb asymmetry is associated with reduced sprinting performance, and the authors conclude that reducing asymmetry does not necessarily need to be a central focus of training when the goal is speed. Small fluctuations in asymmetry over time, as shown in the one longitudinal study included, likely reflect normal neuromuscular variability rather than mechanical inefficiency. That said, asymmetry reduction may still be worthwhile for other reasons, such as injury prevention, and targeted strengthening of the weaker limb has been shown to increase its capacity and reduce side-to-side differences. If practitioners do choose to address asymmetry, the evidence suggests they should do so with task-specific, metric-specific and muscle-specific assessments rather than blanket correction programs.

What remains genuinely unresolved is the question of cause and effect. With nine of the ten studies using cross-sectional designs, it is impossible to say whether asymmetry shapes sprint performance, whether sprinting shapes asymmetry, or whether both simply reflect each athlete’s unique neuromuscular signature. Exell and colleagues observed that two athletes with comparable kinetic asymmetries produced markedly different step velocities, hinting that sensitivity to asymmetry varies from person to person, with some athletes compensating effectively and others not. The review’s authors call for standardized asymmetry calculations, robust longitudinal designs, and the establishment of task-, metric- and muscle-specific thresholds that account for individual responders and non-responders. Until then, the sprinter’s lopsided stride, once viewed as a flaw to be engineered away, may deserve to be seen for what the evidence now suggests it often is: a harmless quirk of a body built for speed.

Subject of Research: The relationship between inter-limb asymmetry and sprinting performance in sprint athletes

Article Title: The Relationship Between Inter-limb Asymmetry and Sprinting Performance in Sprint Athletes: A Systematic Review

Article References: D’Hondt, J., Chapelle, L., Van Handenhoven, J., Stalmans, J., & Aerenhouts, D. (2026). The Relationship Between Inter-limb Asymmetry and Sprinting Performance in Sprint Athletes: A Systematic Review. Sports Medicine – Open, 12(1), Article 146. https://doi.org/10.1186/s40798-026-01082-1

Image Credits: AI Generated

DOI: 10.1186/s40798-026-01082-1

Keywords: inter-limb asymmetry, sprinting, biomechanics, sports medicine, systematic review, kinematics, kinetics, muscle morphology, athletic performance, sprint athletes, neuromuscular, training