The Institute of Robotics and Biomechanics in Warsaw has released version 4.2 of the Arm Path Dataset, incorporating over 12,000 new motion sequences recorded from 340 participants. The update expands the original collection by 45 percent and introduces synchronized electromyography signals alongside high-resolution optical tracking data. Researchers collected samples across five age groups and multiple occupational backgrounds to improve representation of everyday arm movements in controlled laboratory settings.

Key Technical Improvements

Annotation protocols now include detailed joint-angle trajectories and velocity profiles at 200 frames per second. Each sequence carries standardized metadata describing task type, grip configuration, and external load conditions. The dataset also features corrected calibration files that reduce positional error to under 1.2 millimeters on average. A new validation subset of 1,800 sequences has been independently reviewed by three external laboratories to ensure consistency with international motion-capture standards.

Access procedures remain unchanged, with files available under a Creative Commons Attribution-NonCommercial license through the institute’s repository. Documentation has been updated to include Python and MATLAB scripts for basic filtering and feature extraction. Users can download either the full archive or task-specific subsets to reduce transfer times.

Impact on Ongoing Studies

Early adopters report improved performance of machine-learning models trained on the expanded collection, particularly for predicting fatigue onset during repetitive reaching tasks. Several European projects examining upper-limb prosthetics have already integrated the new recordings into their training pipelines. The institute plans quarterly incremental releases to maintain currency with emerging research questions in human-robot collaboration.

Future versions will incorporate data from wearable inertial sensors and additional clinical populations. The current release is expected to support comparative analyses across different demographic cohorts and to facilitate cross-validation of algorithms developed on earlier dataset editions.