Robotic Neck Motion Video Dataset

Preview

Access a sample playlist of videos here: https://www.youtube.com/watch?v=_wOgNZWWoEE&list=PLrP4k_0quIDOl6A6J49_GtRlZbLaV964v&index=1

Description

This dataset contains 84 video recordings of a robotic head and neck mounted on a UR3 manipulator performing controlled neck movements. The dataset was developed to investigate the relationship between neck kinematics and the visual information generated by robot motion during human–robot interaction.

The recordings capture three fundamental neck motion types corresponding to the primary rotational degrees of freedom of the human cervical spine:

Roll (lateral head tilt)

Pitch (head nodding)

Yaw (head shaking)

Each motion is performed under systematically varied kinematic parameters, enabling quantitative analysis of how movement amplitude, acceleration, and repetition influence the visual salience and information content of robotic neck behaviour.

Dataset Composition

The dataset consists of videos generated from combinations of:

Parameter
Values

Motion axis
Roll, Pitch, Yaw

Angular displacement
5°, 10°

Acceleration
0.5, 2.0 m/s²

Repetition frequency
1, 2, 3 repetitions

All videos were recorded using a fixed camera positioned in front of the robotic head under consistent experimental conditions. 

There are two versions of the dataset available: a compressed (207MB) and a non-compressed (3.52 GB) dataset. 

File Naming Convention

Each video file follows the format:

(motion_axis)_(angle)_(acceleration)_(frequency).mp4

Examples:

roll_10_2.0_3.mp4
pitch_5_0.5_1.mp4
yaw_10_0.5_2.mp4

where:

motion_axis ∈ {roll, pitch, yaw}

angle ∈ {5, 10} degrees

acceleration ∈ {0.5, 2.0} m/s²

frequency ∈ {1, 2, 3}

For example, the file

roll_10_2.0_3.mp4

contains a roll motion with a maximum angular displacement of 10°, executed with an acceleration of 2.0 m/s² and repeated three times.

Visual Information Analysis

The dataset was designed to support the measurement of visual information generated by robot neck movements. However it can also be used for surveys and determining future robotic designs as well. 

Applications

Potential applications of the dataset include:

Human–robot interaction research

Analysis of expressive robotic motion

Embodied information and communication studies

Motion perception and visual salience modelling

Robot behaviour generation and evaluation

Benchmarking information-theoretic measures of robot movement

Citation

If you use this dataset in your research, please cite the associated publication describing the embodied information framework and visual entropy analysis methodology. This will be acceptance upon the acceptance of the publication that carries more information regarding the dataset.