Hand exoskeletons for robot learning and rehabilitation need finger joint angles, but self-aligning designs that report them usually compute them through a model calibrated to each wearer. A few finger devices avoid the model by letting each segment slide along its phalanx without rotating on it. We give this condition, rotation locking, a general analysis. In a rigid planar model the encoder between two segments reads the joint angle plus a constant for any phalanx length, standoff or axis migration within the slides’ travel, and torque applied between the segments reaches the joint as a pure couple through frictionless slides. Each slide must travel h·tan(θ/2), segments spanned by two joints must telescope, and the straight finger is singular. In 5,000 simulated joints with one reference pose each, one rotation-locked encoder erred by a median of 1.5° r.m.s., against 2.1° for one encoder on a generic revolute–prismatic–revolute chain. TAKTO ONE, an open-source four-finger exoskeleton with twelve joint encoders, is built on these rules and works on different hands; its accuracy against a reference is not yet measured. Files: the preprint (PDF), the analysis and figure code with its data (ZIP), Video S1 (the device's twin in augmented reality, moved by its encoders) and Video S2 (the device worn). Design files under open licences: https://github.com/molanocortes/takto-one. Project page: https://takto.one Version 2 (2026-10-19): new title; cites DEXOP and DexEXO; Eq. 4 simplified and an unused symbol removed; AI-use statement reworded. The results, figures and data are unchanged.
No takes yet. Share an insight, caveat, or question.
Molano et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: