Related Experiment Video
Updated: May 14, 2026
07:14
Tracking Infiltration Front Depth Using Time-lapse Multi-offset Gathers Collected with Array Antenna Ground Penetrating Radar
Published on: May 1, 2018
A Dataset for Victim Detection in Search-and-Rescue Operations Using Robot-Mounted UWB-Radar Sensors.
Antonios-Periklis Michalopoulos1,2, Efstratios N Paliodimos1, Fotios Papadopoulos1
1Department of Electrical and Electronics Engineering, University of West Attica, Athens, Greece.
Scientific Data
|May 12, 2026
Summary
This study introduces a new dataset for robot-mounted Ultra-Wideband radar in Search-and-Rescue (SAR) operations. It enables AI development for detecting victims behind obstacles, achieving high accuracy.
Area of Science:
- Robotics and Sensor Technology
- Artificial Intelligence in Disaster Response
- Radar Sensing for Search and Rescue
Background:
- Search-and-Rescue (SAR) operations require effective victim detection, especially in occluded environments where conventional sensors fail.
- Robotic platforms offer enhanced capabilities for disaster response in hazardous conditions.
- Ultra-Wideband (UWB) radar shows promise for penetrating obstacles, overcoming limitations of thermal and optical sensors.
Purpose of the Study:
- To introduce a novel dataset of victim measurements using a robot-mounted UWB radar sensor.
- To emulate realistic SAR scenarios within buildings for victim detection and localization.
- To provide a foundational resource for developing advanced AI/ML models for SAR operations.
Main Methods:
- Collected a comprehensive dataset of victim measurements using a robot-mounted UWB radar sensor.
- Designed dataset scenarios to simulate indoor SAR operations with victim detection challenges.
- Developed and evaluated Convolutional Neural Network (CNN) and XGBoost models for victim detection.
- Implemented a rule-based method for victim position estimation.
Main Results:
- The dataset is the first to capture robot movement during SAR operations with UWB radar from multiple orientations.
- CNN and XGBoost models achieved F1-scores of 78% and 83% for victim detection, respectively.
- The rule-based position estimation method achieved a Mean Absolute Error (MAE) of 0.49m.
Conclusions:
- The presented dataset is valuable for advancing AI/ML in SAR by providing realistic UWB radar measurements.
- The dataset supports the development of robust victim detection and localization algorithms for robotic SAR.
- This work highlights the potential of UWB radar and robotic platforms in improving SAR efficiency and effectiveness.
