International Journal of Science, Engineering and TechnologyAuthors: Almutasim Billa Alanazi, Abdullah Bamufleh, Mohammed Arnaout
Abstract: Warehouse automation is increasingly important for improving operational efficiency, reducing sorting errors, and supporting high-throughput logistics. This paper presents the design, implementation, and experimental evaluation of an AI-based autonomous warehouse robot car for real-time color-based sorting of uniformly sized boxes in an indoor warehouse environment. The robot integrates a dual-camera perception architecture, an ONNX-optimized YOLO object-detection model, ArUco marker-based localization, a Raspberry Pi 5 embedded processor, four Mecanum wheels for omnidirectional mobility, ultrasonic sensing for final approach control, a servo-driven gripper, and a Flask-based real-time monitoring interface. Autonomous task execution is organized using a finite-state machine that sequences Search, Align, Approach, Pick, Deliver, and Exit states, while a deterministic priority rule (Red > Blue > Green) defines the order of box handling. Experimental testing was performed in an approximately 3 m × 3 m simulated warehouse with multiple runs and varying numbers of red, blue, and green boxes. The reported results showed object-detection accuracy of 85%–95%, alignment accuracy of approximately 92%, pick-and-place success of 91%–95% with more than 95% success in most test conditions, and delivery accuracy of approximately 97%. The average detection-to-delivery time was approximately 20 s per box, system latency was below 100 ms, and the priority-sorting logic achieved 100% correctness with no cross-color delivery errors. The findings demonstrate the effectiveness of integrating edge AI, marker-based localization, omnidirectional locomotion, robotic manipulation, and real-time monitoring on a low-cost embedded platform.
