Development of a robot localization and environment mapping system

Abstract

The intent of this research is to develop a robust, efficient, self-contained localization module for use in a robotic liquefied petroleum gas (LPG) tank inspection system. Inspecting large LPG tanks for defects is difficult, expensive and energy intensive. Replacing the human inspectors with a robotic inspection system will make the inspection process faster, less expensive, more reliable and safer. The sensing platform designed in this work can collect data about the environment and track the robotic inspection platform, recording the defect locations. It consists of a two axis gimbaled sensing platform with a single point distance sensor placed in the manhole of the tank. A collection of algorithms were developed to use in conjunction with the sensing platform to collect and process the 3D data into a map of the environment. The algorithm's main feature is a robust and efficient method of segmenting and fitting data to a right capped cylinder that is faster and more robust to noise than current methods. The improved performance comes from a unique combination of object shape knowledge, the Gauss image and 3D histogram techniques which achieve accurate segmentation without iteration. The hardware and software were demonstrated to function robustly in a noisy environment. The unique ability of the system to work in an LPG tank allows it to be integrated into a robotic inspection system that can remove the majority of the cost and risk associated with LPG tank inspection.