How ayaancdry/agribot works
Project type and purpose
• This is a ROS-based robotics workstation (“Agribot”): a full-stack robotics software stack combining simulation (Gazebo/ROS), perception (YOLO object detection), and control (pot follower).
• The top-level “src” hosts multiple ROS2 packages:
• sim_bot (robot simulation),
• sim_lobby_classic (environment/world),
• yolo_msgs (custom ROS message definitions),
• yolo_ros & yolo_bringup (YOLO inference nodes & launch),
• pot_follower (control node reacting to detections).
File structure
• src/… directories correspond to ROS2 packages, each with package.xml, setup.py, and Python/CMake code under src.
• build/ and install/ are Colcon workspaces artifacts.
• custom_yolo_models/ holds ML weights.
• runs/ and train_pot_detector.py for offline training.
• ROS launch files in each package’s launch/ or share/…/launch define how to spin up nodes and Gazebo worlds.
• yolo_msgs defines the message schema for detection data shared across nodes.
README
• Minimal description indicates this is the “Agribot” workstation. We infer the stack runs end-to-end: simulated sensors → YOLO detection → control logic → simulated actuation.
Elements
• “Gazebo Simulator” (with Camera & Robot model plugins).
• “sim_lobby_classic” world loader.
• YOLO Processing Nodes (yolo_node, detect_3d_node, tracking_node).
• “yolo_msgs” interface block.
• “pot_follower_node”.
• “custom_yolo_models” weights.
• Image → YOLO Nodes.
• DetectionArray → pot_follower.
• Twist → Gazebo for robot motion.
• Green for simulation components.
• Blue for perception components.
• Orange for control components.