The Motions Dataset was released in 2023 by researchers from the Max Planck Institute for Intelligent Systems.Safe & Accurate at Speed with Tendons: A Robot Arm for Exploring Dynamic Motion", and has been accepted by RSS 2024. This dataset contains long-term dynamic motion data collected from a newly designed 4-DOF pneumatic muscle-driven bionic arm (Pamy2), with a data collection period of approximately 3.5 weeks. The data includes target pressure motion generated by random multi-frequency signals, as well as repeated fixed target pressure motion, which helps to evaluate the repeatability and performance of the system.

The dataset contains 6 zip files, each of which contains about 40 "df" files, and each "df" file contains a pandas data frame. Each data frame can be read through the pandas library, each row corresponds to an observation of the robot state, and the columns of the data frame have clear names.

The robot's motion is annotated into different behavior segments, which are provided as intervals of iteration numbers ("iteration" is one of the data frame columns). These intervals are provided as attributes of the data frame. There are 5 annotation categories:

1. Random Motion: Open-loop motion generated by random multi-frequency signals of target pressure at different frequencies.
2. Fast Fixed Motion: An open loop motion consisting of a series of fixed target pressures with a time interval of 0.7 seconds between each target pressure.
3. Slow Fixed Motion: An open loop motion consisting of a series of fixed target pressures with a 2 second interval between each target pressure.
4. Random Pressure Move: Movement generated by applying random target pressure.
5. Reset: Reset the motion by moving to medium pressure, then to minimum pressure, and back to medium pressure; the final position may be suitable for evaluating the repeatability of the system.