----Robotic Arm Dataset Readme File----

Underpinning Research Article: Task-oriented Prediction and Communication Co-Design for Haptic Communications
Authors: Burak Kizilkaya, Changyang She, Guodong Zhao, Muhammad Ali Imran
Contact: b.kizilkaya.1@research.gla.ac.uk, guodong.zhao@glasgow.ac.uk

Dataset Description: The real-world trajectory data samples are collected from our teleoperation testbed at University of Glasgow. The robotic arm is controlled by a human user to finish three types of tasks:
1) Pushing a box: Push a small box from the starting point to the end point along a given routine
2) Grouping items with different colors: Move items with the same color to the same area
3) Writing symbols: Write symbols by controlling the robotic arm. 
For each task, there is a dedicated task folder and there are multiple experiments in each task folder. Each experiment dataset is provided in .csv format. Data are sampled from the robot in every 1 millisecond. In each .csv file, there are 28 columns.
Columns:
[1] time
[2-8] 7 joint angles
[9-15] 7 joint velocities
[16-22] 7 joint accelerations
[23-25] 3 external forces in x, y, z directions
[26-28] 3 external torques in x, y, z directions