TAVIB: Tactile sensing time series data for robotic object interaction

TAVIB (Tactile Vibration) is a curated collection of tactile time series for robotic object interaction. It contains 9,536 CSV recording files sampled at 8 kHz: 1,750 experimental active-touch recordings of manufactured objects, 1,750 simulation files representing 35 unique simulated responses, 6,000 active-touch recordings from 30 strawberries measured over four days, and 36 slippage and sliding recordings covering three materials, four orientations, and three nominal speeds.

The active-touch subsets contain transmitter and receiver vibration signals. The slippage subset contains six accelerometer channels and two encoder-derived velocity channels. The archive includes a README describing formats and columns, recording-level metadata, mappings from original to release paths, and a SHA-256 checksum manifest.

Provenance: The active-touch experimental, simulation, and strawberry data were previously deposited by Danissa Sandykbayeva and Valeriya Zhanuzakova in the OSF dataset 10.17605/OSF.IO/G4AE8 and described in Sandykbayeva et al., Scientific Reports (2026). The slippage and sliding recordings originate from the experiments described in Massalim et al., Sensors (2020). TAVIB combines these sources using a unified directory structure and recording-level metadata; it does not represent a new data acquisition.

Reuse: The recordings support tactile signal processing and material or contact-motion classification. They may support research on tactile representations for vision-language-action models, but do not constitute synchronized vision-language-action demonstrations.

Repeated contents: Each simulated response is retained under 50 source-derived file paths. The aluminum slippage recordings at 0 degrees and nominal medium and high speeds have identical contents. File counts should therefore not be interpreted as counts of independent observations.