Design of an automated stock-taking system based on unmanned aerial vehicles

Graphical/Tabular Abstract In this study, a UAV-based system that automates stock-taking process in a warehouse environment is proposed. This system moves autonomously in the warehouse and captures the barcodes of the products from the images collected by the camera. The captured barcodes are decoded online and the product information is recorded in the inventory file. Thus, a ground station computer can access this constantly updated inventory file on the UAV via Wi-Fi. The proposed system approach is represented in Figure A. Figure A. The general picture of proposed stock-taking system Purpose: The UAV-based stock-taking system proposed in this study basically aims to solve three main problems. (1) The indoor positioning problem is still a major challenge. In order for the UAV-based automatic stock-taking system to operate autonomously in an indoor environment, the problem of indoor 3D positioning problem is solved. (2) Images taken with the camera on the UAV, which moves autonomously and scans all shelves, are processed online, barcodes are captured and stock-taking is made. (3) In the warehouse environment where commercial activity continues, the obstacle avoidance function that will prevent collisions with people is operated. Theory and Methods: The solution methods of the three basic problems mentioned above were as follows: (1) Indoor positioning is provided by the installation of an UWB-based positioning system in a laboratory environment. Four UWB transmitter modules are placed in certain positions in the laboratory to produce the positioning of the UAV. (2) Once the positioning solution is provided, the UAV autonomously scans the shelves using the greedy search method and collects image data. These data are processed using the OpenCV framework and barcodes are detected. (3) When the UAV encounters a dynamic obstacle while performing its counting function, it detects the obstacle with the data it receives in the LiDAR and makes an avoidance maneuver. Results: (1) Indoor positioning problem is solved precisely with an error margin of 4 cm. (2) All of the barcodes produced in the Code-128 standard are successfully detected and the stock count is done correctly. (3) When there is an obstacle approaching more than 1m to the UAV, the UAV successfully made an avoidance maneuver and continued to operate safely. Conclusion: All subsystems of the proposed UAV-based automatic stock counting system approach is verified by laboratory tests. The final flight tests are shown that it successfully counts the stock with all its functions.