Autonomous Date-Palm Harvesting UGV

Designed and delivered a full-stack autonomous agricultural robot capable of detecting, navigating to, and harvesting date-palm fruit clusters across large open farm fields.

An NVIDIA Jetson Orin runs ROS2 as the middleware backbone, tying together perception, planning, and actuation. YOLOv7 + OpenCV handles real-time fruit detection — identifying ripe clusters by color, shape, and canopy position — while RRT* path planning computes optimal collision-free routes between palm rows. A dedicated obstacle avoidance layer triggers local replanning mid-traverse when obstructions are encountered. The entire navigation and harvesting stack was validated in Gazebo simulation before field deployment.

ATmega2560 microcontroller handles all low-level hardware in real time — motors, linear actuators, solenoid valves, and relays — receiving commands from the Jetson over I2C, with precise actuation sequences ensuring safe grip, cut, and retract cycles. The ~800 kg platform runs on a LiFePO4 pack (600 Ah) delivering ~5 hours of field operation, with a wireless PS2 controller added in revision for manual override.