architecture.md

Software architecture

The software combines global navigation with local vision-based control.

flowchart LR
    Camera[Pi Camera] --> Detector[YOLOv8 target detector]
    Detector --> Color[HSV color verification]
    Color --> Range[Monocular distance estimate]
    Range --> Controller[Yaw / altitude / distance control]
    Controller --> MAVLink[MAVLink velocity commands]
    MAVLink --> FC[ArduPilot flight controller]
    FC --> Telemetry[GPS / heading / altitude telemetry]
    Telemetry --> Mission[Mission state logic]
    Mission --> Controller
    Mission --> Servo[Manipulator servo commands]
    Camera --> Stream[MJPEG monitoring stream]

Loading

Main mission stages

1. Connect to the flight controller and camera.
2. Navigate through configured GPS points.
3. Descend to the visual-search altitude.
4. Detect and verify the selected target.
5. Estimate target distance from image geometry.
6. Generate body-frame velocity, yaw-rate, and altitude corrections.
7. Recover toward the configured target location if visual contact is lost.
8. Actuate the manipulator, retreat, return, and land.

The flight parameters, control gains, camera calibration, and servo PWM values are platform-specific and must be validated for each vehicle.