ROS-based Control System for Localization and Object Identification of Indoor Self-driving Mobile Robot

In this study, an indoor self-driving automated guided vehicle that recognizes objects in a logistics center and transports them to their destinations is introduced. The robot must be aware of its location while it moves and must be able to recognize its surroundings in real time to operate in the self-driving mode. The control system is composed of a robot operating system (ROS) with OpenCR used as a lower controller for motor control and other sensor processing. The robot position is estimated by combining the adaptive Monte Carlo localization algorithm (AMCL) with odometry using an inertial measurement unit sensor and an encoder, and mapping and navigation are performed by combining it with LiDAR. In addition, box classification is conducted through you-only-look-once (YOLO) object detection. Consequently, we implement a self-driving robot that sets its own target point through box classification and avoids obstacles in real time by recognizing obstacles with LiDAR.