BEE 3900: Bio Robot

This project is an autonomous crop-harvesting robot developed for BEE 3900 at Cornell University. The robot was designed to navigate a row-planted field, detect crops using computer vision, classify plant condition, and perform pruning or collection with a robotic arm.

The system uses a Raspberry Pi and Arduino-based control architecture. The Raspberry Pi handles high-level tasks such as camera input, image processing, object detection, classification, and robotic arm control. The Arduino manages low-level hardware operations, including stepper motor control, line-following logic, LED signaling, and sensor feedback.

The mobile platform is equipped with four mecanum wheels, enabling omnidirectional motion for precise alignment and positioning. An Elegoo three-channel line tracker supports autonomous navigation along printed field paths, while an IMU and limit switch provide additional feedback for movement and control.

For manipulation, the robot uses a Le Robot SO-101 robotic arm. Arm poses are converted into PWM commands to control the servo motors, allowing the robot to position its claw around target crops and complete the pruning or collection action.

The perception system uses an Innomaker U20CAM-M-1080P camera with LED-assisted illumination. We implemented a computer vision pipeline using YOLOv8 Nano for fast plant detection and MobileNetV3 for crop classification, distinguishing healthy dark green crops from unhealthy light green crops. Classification results were saved to CSV files for later analysis and offloading.

This project integrated embedded systems, robotic motion control, computer vision, and machine learning into a compact autonomous agricultural robot. It demonstrated a full workflow from field navigation and crop detection to robotic arm positioning and physical crop collection.

Poster

Gallery