Proprioceptive estimation of forearm roll (q5) — dataset (Experiments 2 & 3) Code: https://github.com/TaISLab/proprioceptive_limb_roll Method/thesis: Castro Ochoa, R. "Sistema multisensorial para la interacción física humano-robot." TFM, Universidad de Málaga, 2026. Contents - `exp2_discrete/exp2_discrete_anonymized.csv` — 97 discrete captures, 8 anonymised participants (9 sessions — one participant has both a right- and a left-arm session). Columns: `id`, `participant`, `q5_gt` (deg, IMU ground truth), `dedo1_q1`…`dedo4_q2` (deg, phalange encoder angles). See the code repo's `data/README.md` for the full column description and the `dedoN` → pinch/side mapping. - `exp2_discrete/participants_meta.csv` — per-participant `sex` and `arm_tested`. **Forearm section semi-axes (`a`, `b`) were not collected for these participants** — not included, not a placeholder. - `exp3_continuous/exp3_continuous_sweep01_raw.csv` — one continuous pronation→supination sweep, raw rosbag-style export (one row per ROS message; sparse/misaligned columns, not a tidy table). Includes IMU ground truth (`/roll_angle_x/data`), per-pinch local q5 estimates (`/tactile/q5_pinza_1/angle`, `/tactile/q5_pinza_2/angle`), the forearm-axis reconstruction topics (`/tactile/info/...`), and the precomputed `q5_inv` / `error_abs` columns (sign-converted estimate and its absolute error against ground truth, at sample resolution). Ethics & consent Collected under signed informed consent with guaranteed participant anonymisation, approved by the "Comité Ético de Experimentación de la Universidad de Málaga" (CEUMA), registration number 7-2023-H. No names, dates of birth, or other identifying data are included; participant codes (`P01`…) are unrelated to any real name and the correspondence is held privately by the research team, not distributed with this dataset. Funding Project PID2021-127221OB-I00 (CONCERTO), funded by MICIU/AEI/10.13039/501100011033/FEDER, UE. Citation Castro Ochoa, R., Gómez de Gabriel, J. M., Urdiales, C., de Cózar, Ó. and Blázquez, B. (2026). *Proprioceptive estimation of forearm roll during robotic grasping — dataset.* TaISLab, Universidad de Málaga.