Semi-autonomous computer vision-based rockbreaker

Abstract

Automating rockbreaker manipulation improves the safety and performance of mining site operations. The goal of this project is to design and develop a semi-autonomous rockbreaker. This research contributed to both the industrial and scientific aspects of this problem. For the industrial part, we implemented a complete semi-autonomous rockbreaker’s chain. Our proposed system will solve teleportation challenges. Also, it will fulfill some of the requirements of a fully autonomous rockbreaker. Besides, it can be used for future research toward a final fully autonomous rockbreaker. We proposed a heuristic path planner algorithm that can compete with a trained operator in terms of both speed and accuracy of reaching the target position. For the scientific part, we proposed a camera-based arm tracking algorithm that overcomes the shortcomings of the current non-camera-based approach. Moreover, it decreases both installation and maintenance costs. Furthermore, it is more stable than the current approach as well. Using our industrial product, we suggested a method to automatically generate labelled data to train our camera-based arm tracking method.