What Contact Can Tell

A new review maps how robotics and artificial intelligence are converging in tribology: the study of friction, wear and lubrication. The authors describe automated sample preparation, high-throughput tribometers and multimodal in-situ sensing as the physical layer; supervised learning, physics-informed models and Bayesian optimization form the analytical layer. Joined in a closed loop, a test rig could measure contact, interpret the result and select the next experiment. The paper is a synthesis, not a newly demonstrated autonomous laboratory. It identifies partial implementations and unresolved barriers including inconsistent datasets, model transfer between materials and hardware-software integration. The useful shift is architectural: wear testing can move from repeated observation toward an instrument that reads its own evidence and changes course within explicit limits.

Researchers proposed a data-driven method for generating shoe-insole outlines across sizes without assuming that every part scales uniformly. The workflow extracts two-dimensional contours, represents them as B-splines and predicts local control-point movement along tangent and normal directions. In an eight-fold leave-one-insole-out evaluation, the method reported a mean Hausdorff distance of 4.73 millimeters, a Chamfer distance of 1.76 millimeters and intersection-over-union of 0.929. It significantly outperformed a no-clustering version and a PCA-based kernel-ridge baseline, but did not significantly beat ratio scaling, Gaussian-process or thin-plate-spline baselines; PCA-linear was numerically better on Hausdorff distance without a significant difference. This is standard-size digital grading, not personalized fit or a validated manufacturing pipeline. Its value is a more local account of where shape actually changes.

A flexible pressure sensor uses a sponge whose pore size changes through its thickness. Sugar particles serve as a sacrificial template inside a conductive polydimethylsiloxane composite, producing larger pores for compressibility and smaller pores for mechanical stability. The authors report sensitivity of 5.6 per kilopascal, a zero-to-350-kilopascal detection range and strain tolerance up to 70 percent. Demonstrations include human-motion detection, foot-pressure mapping and Morse-code input through pressure pulses. Those examples establish laboratory versatility, not long-term wear, washability or clinical reliability. The design principle is more durable than the demonstrations: one graded physical structure can translate both light and heavy contact without requiring the person to switch sensing modes.

An anisotropic conductive hydrogel approaches brain-computer interface sensing as a contact problem. Directional freezing during copolymerization creates aligned vertical microchannels for ion transport alongside an interconnected nanoporous network for electronic conduction. The authors report a tissue-matched compressive modulus, high extensibility, skin adhesion and rapid self-healing. Preliminary P300 and steady-state visual-evoked-potential tests produced signal quality comparable with conventional wet and dry electrode benchmarks. The published abstract does not establish large-cohort performance, repeated daily use or how healing affects calibration over time. Still, the work aligns electrical pathways, mechanical compliance and repair within one material architecture. A wearable electrode can adapt at the skin rather than asking straps, adhesives or setup rituals to compensate for every movement.

A musculoskeletal-robotics study combines parametric design with flexible 3D printing to produce muscle-, tendon-, ligament- and cushion-like components from one thermoplastic polyurethane. Lattice density and branch thickness tune mechanical behavior by body region, while ligament-based joint constraints and tendon-driven actuation support coordinated multijoint leg motion at life scale. Under external suspension and partial body-weight support, the prototype constrained torso roll, pitch and yaw between minus 20 and plus 20 degrees. The result is a fabrication framework, not a free-standing humanoid, and embedded sensing remains a proposed extension. What changes is the relationship between geometry and behavior: compliance becomes a design variable that can be generated and revised systematically, reducing one-off mechanical customization.