Anomaly Detection Scenarios for Turtlebot3 Burger and Panda Emika Franka

This dataset contains experimental logs collected for the study"Data-Light Hybrid Anomaly Detection for Robotic Systems Using Semantic and Message-Count Features"conducted at Ben-Gurion University of the Negev.

The data includes real-world and simulated runs from two robotic platforms:

TurtleBot3 Burger — navigation tasks

Franka Emika Panda — manipulation tasks

Each experiment includes normal runs and multiple anomaly scenarios such as:

sensor degradation

control disturbances

network degradation

collision and environment interference

adversarial velocity / command perturbations

Dataset Contents

For each robot and scenario, we provide:

Raw ROS log files (.bag or extracted structured logs)

Pre-processed feature logs used for analysis

Semantic features (robot state variables: position, velocity, joint states)

Message-count features (count histograms of ROS messages per time window)

Format

Timestamp-indexed log files

CSV and structured feature matrices

Metadata files describing scenario type and label (normal/anomalous)

Purpose

This dataset supports research in:

anomaly detection in robotics

robust robot performance monitoring

ROS-based behavior analysis and cyber-physical safety

hybrid semantic + communication-level modeling

It enables reproducible evaluation of data-light anomaly detection methods under diverse anomaly types and realistic robotic behavior variability.

Usage

The dataset is suitable for researchers exploring anomaly detection, robotics safety, ROS analytics, or heterogeneous feature modeling.