Development and Performance Analysis of a Human Detection Robot Using YOLOv8 and PWM-Based Speed Control

This paper presents the design and performance evaluation of a human detection robot using the YOLOv8 model and the COCO dataset for object recognition. The robot is equipped with a Pi camera, Raspberry Pi, four GM25 13CPR motors, an L298 motor driver, and a buck converter, ensuring efficient operation in real-time environments. The human detection accuracy was evaluated at different distances, achieving 99% at 2 feet, 98% at 15 feet, and 96% at 25 feet, demonstrating the effectiveness of the YOLOv8 model in varying conditions.The robot's movement is controlled using a PWM-based speed control technique, where the DC motors operate at different duty cycles. Experimental results show variations in speed accuracy, with error percentages of 7.6% at 20% duty cycle, 5.8% at 40%, 5.1% at 60%, 4.8% at 80%, and 3.8% at 100% duty cycle. These results indicate that higher duty cycles lead to improved speed accuracy, minimizing the deviation from the desired speed. The study highlights the integration of YOLOv8 for object detection and PWM for precise motor control, making the system suitable for applications in autonomous navigation, surveillance, and security.