Object Classification by a 7-DOF Robotic Arm Using MobileNet-SSD and Feedback Position Mapping

In this paper, a 7-Degrees of Freedom (DOF) robotic arm with a gripper and MG996R servo motors is utilized for object classification tasks, integrated with MobileNet-SSD, an object detection model, without the use of kinematic equations. To implement this control and manipulation system, precise tasks are executed using an ATmega2560 microcontroller, by using a deep learning-based computer vision model employed for object detection, and positioning is achieved through pre-recorded servo angles based on feedback position mapping (III-D). The aim of this study is to precisely control a high-Degrees of Freedom (DOF) robotic arm without relying on complex kinematics and control algorithms, utilizing a fast and accurate object detection model for tasks such as object classification in manufacturing, laboratory research, warehouse logistics, and medical assistance, all while operating on microcontrollers with limited processing power. Additionally, the proposed approach has been experimentally tested and verified, with results confirming its effectiveness and demonstrating that the integration can be successfully applied to more sophisticated automation solution.