DevinLite

Autonomous USV (Unmanned Surface Vehicle) control system for RoboBoat 2025 competition using ZED2i camera and YOLO object detection.

Features

• Real-time buoy detection and classification
• Autonomous navigation between red and green buoys
• Hazard avoidance (yellow and black buoys)
• Multiple navigation scenarios:
  • Single buoy detection (red/green)
  • Hazard buoy detection
  • No-detection search pattern
  • Triangle-based path planning

Requirements

• ZED2i camera
• CUDA 12.6
• A100 GPU
• Python dependencies:
  • pyzed
  • opencv-python
  • numpy
  • ultralytics (YOLO)
  • supervision

Motor Control

PWM values for different motions:

• Stop: 1500 PWM
• Forward (straight): 1560 PWM
• Turn right: Right motor 1500 PWM, Left motor 1560-1570 PWM
• Turn left: Left motor 1500 PWM, Right motor 1560-1570 PWM
• Reverse: 1450 PWM

Usage

Run the main script:

python main.py

The script will initialize the ZED camera, load the YOLO model, and begin autonomous navigation.