Related Experiment Video
Updated: Jul 5, 2025

Tracking Infiltration Front Depth Using Time-lapse Multi-offset Gathers Collected with Array Antenna Ground Penetrating Radar
Published on: May 1, 2018
Research on Time Series-Based Pipeline Ground Penetrating Radar Calibration Angle Prediction Algorithm
Maoxuan Xu1, Feng Yang1, Yuanjin Fang1
1School of Mechanical Electronic and Information Engineering, China University of Mining and Technology (Beijing), Beijing 100083, China.
A deep learning model using Long Short-Term Memory (LSTM) effectively predicts guiding wheel angles and torque for pipeline robots. This enhances ground-penetrating radar accuracy in detecting underground pipeline defects.
Area of Science:
- Geophysics
- Robotics
- Artificial Intelligence
Background:
- Underground infrastructure security relies on accurate detection of defects within pipelines.
- Pipeline robots navigating underground drainage systems experience posture deviations due to various factors.
- Precise spatial positioning of ground-penetrating radar (GPR) antennas is crucial for defect detection.
Purpose of the Study:
- To develop an intelligent control system for correcting pipeline robot posture.
- To propose a time-series-based algorithm for predicting guiding wheel correction angles and torque.
- To leverage deep learning for enhancing the accuracy of GPR defect localization.
Main Methods:
- A wheeled pipeline robot equipped with a guiding wheel was utilized for posture correction.
- A Long Short-Term Memory (LSTM) deep learning model was employed for predicting correction angles and torque.
- The performance of the LSTM model was compared against an Autoregressive Integrated Moving Average (ARIMA) model.
Main Results:
- The LSTM model demonstrated reduced mean absolute error (MAE) by 4.11° for angles and 8.25 N·m for torque.
- Mean squared error (MSE) for predicted angles and torques decreased by 10.66% and 7.27%, respectively.
- Experimental results showed an average correction speed of 5 seconds and an angular error of ±1°.
Conclusions:
- The LSTM-based prediction model effectively corrects pipeline robot posture in real-time with high accuracy.
- This intelligent attitude correction significantly enhances the precision of GPR antennas in locating pipeline defects.
- The proposed method offers a robust solution for improving underground space security through advanced robotic GPR systems.
Related Concept Videos
Taping Over Different Ground Profiles
Common Leveling Mistakes and Errors
Distance Corrections
Adjusting a Traverse
Design Example: Traverse Angle Computations
Influence of Earth's Curvature and Atmospheric Refraction on Leveling

