Design and Development of a Multifunctional Smart Agricultural Robot (SARDOG)

Faculty

Hovannes Kulhandjian, Ph.D., Associate Professor, Department of Electrical and Computer Engineering, California State University, Fresno

Student Researchers

  • Undergraduate students contributed to AI-based strawberry harvesting using a robotic arm as part of their senior design project
  • Student contributors include Nicholas Amely, Devin Rocha, and Blake Bennett, who received awards at the Fresno State Innovation & Entrepreneurship Summit
  • High school students participated in the Finish-in-Five summer research experience, contributing to pollination robot design

Project Description

The SARDOG project integrates artificial intelligence, robotics, and automation into agriculture by developing multifunctional robotic systems for fruit harvesting, weed elimination, spraying, and pollination. A central component is an AI-powered fruit harvesting system that employs a six-degrees-of-freedom robotic arm mounted on the SARDOG platform. The arm is controlled by a Raspberry Pi 5 and an NVIDIA Jetson Orin, which communicate via GPIO relays to synchronize motion control with real-time fruit detection. The system uses a YOLOv8 deep learning model trained on a custom dataset of eight tree fruit types. This model provides real-time object detection, localization, and classification with up to 88 percent validation accuracy after 25 training epochs. Bounding box coordinates generated by the model are processed to align the robotic arm with detected fruit. An adaptive electronic gripper secures the fruit without the need for bulky pneumatic systems, reducing power consumption while maintaining harvest quality. The robotic arm’s kinematics are managed through Python scripts on the Raspberry Pi, with motion planning simulated in ROS2 RViz. This relative angular control method allows the arm to avoid branches and obstacles during harvesting. Experimental testing showed the system could execute pulling and twisting motions to successfully harvest fruit types such as oranges and plums without damage. Integration into the SARDOG platform involved structural modifications to improve orchard accessibility and ensure stable mounting of hardware.

Together, these innovations allow SARDOG to serve as a modular, AI-driven agricultural robot capable of performing precision harvesting and adaptable to different crops and orchard conditions.

Why Is This Research Important/Relevant

This research addresses critical challenges in agriculture by reducing labor needs, improving efficiency, and enhancing sustainability. By combining robotics with AI, the project supports precision farming and advances automation technologies that can be adopted by growers to improve yield and resource management. It also provides students with practical training in cutting-edge areas of AI, robotics, and agricultural technology.

Research Outcomes/Results

  • Developed and tested prototypes for robotic strawberry harvesting, weed elimination, fruit harvesting, spraying, and pollination

  • Advanced prototyping with integration of 3D vision, YOLOv8-based fruit detection, and soft grippers for real-time harvesting