Mapping Pipelines and Simultaneous Localization for Petrochemical Industry Robots

TL;DR

This paper demonstrates the feasibility of using ROS-based SLAM with sensor fusion for autonomous pipeline inspection robots in complex environments, combining simulation and physical testing to improve mapping accuracy.

Contribution

It introduces a combined simulation and physical implementation of a SLAM-capable robot for pipeline inspection, utilizing sensor fusion to enhance odometry without wheel encoders.

Findings

Sensor fusion improved odometry accuracy.

Simulated and physical robot mapping showed reasonable correspondence.

SLAM is feasible for pipeline inspection with real-world complexities.

Abstract

Inspecting petrochemical pipelines is challenging due to hazardous materials, narrow diameters, and inaccessible locations. Mobile robots are promising for autonomous pipeline inspection and mapping. This project aimed to simulate and implement a robot capable of simultaneous localization and mapping (SLAM) in an indoor maze-like environment representing simplified pipelines. The approach involved simulating a differential drive robot in Gazebo/ROS, equipping it with sensors, implementing SLAM using mapping, and path planning with move_base. A physical robot was then built and tested by manually driving it in a constructed maze while collecting sensor data and mapping. Sensor fusion of wheel encoders, Kinect camera, and inertial measurement unit (IMU) data was explored to improve odometry and mapping accuracy without encoders. The final map had reasonable correspondence to the true maze despite lacking wheel encoders. In summary, results show the feasibility of using ROS-based SLAM for pipeline inspection if accounting for real-world complexities.