This article describes the development and evaluation of an aerial robotic system for smart inventory of stockpile warehouses. The system was developed to automatically measure piles of different bulk materials, such as phosphorus and potassium compounds, which are stored in bays of different sizes inside a warehouse. This warehouse configuration is very common among fertilizer and animal food industries in Brazil. While an inventory can be executed by a human technician, the insalubrious environment, the imprecision of the manual volume estimation, and the time spent by the technician to access the information motivate the automation of the process. The proposed system uses a multirotor electrical drone that navigates autonomously inside the warehouse while it acquires light detection and ranging (LIDAR) point cloud data. This data is used to build a three-dimensional (3D) model of the environment, which is then processed to identify the stockpiles of material and calculate their volumes.Since the environment is GPS-denied and its characteristics, including symmetry, illumination and texture, do not favor visual-or LIDAR-based localization, a drone navigation strategy that relies on relative positioning with respect to simple structures of the warehouse was developed. This article also presents our approach for autonomous stockpile volume estimation, which was numerically evaluated both in simulation and with real data, yielding in accuracy and precision of about 98%. The results presented in the article show that the aerial system is able to substitute the previously adopted manual procedure, highly increasing its accuracy, repeatably and safety, and drastically reducing its time of execution and cost.

No abstract