Towards a Simple Navigation Strategy for Autonomous Inspection of Ducts and Galleries

Abstract

One of the most significant technological challenges in this current fourth industrial revolution (Industry 4.0) is developing and operationalizing industrial mobile robots capable of performing tasks autonomously. In this sense, one of the most critical activities in industrial environments is to inspect regions that can be dangerous for human operators, such as pipes, caves, dam galleries, and tanks. Those areas can be considered confined environments, which were not designed for continuous human occupation. This paper focuses on extensive ducts and galleries with repeating patterns and a limited number of bifurcations that are common in the mineral industry, and are challenging for robots given the communication limitations and lack of an accurate global localization (GPS-denied). We present a navigation strategy for autonomous inspection that does not rely on an accurate global localization or a previous map. The proposed method based upon maze-solving algorithms is fast and can be computed directly in the onboard robot hardware, using only a planar LiDAR and an IMU. The method’s expected behavior relies on a finite state machine that allows for autonomous decision-making on the fly while using two different control algorithms to navigate inside the tunnel’s center and drive the robot autonomously at the intersections. The proposed method was validated in realistic simulations on representative virtual scenarios of pipes and galleries using the CoppeliaSim simulator and ROS, demonstrating the strategy feasibility.