Skip to content

infrastructure.md

Latest commit

 

History

History
145 lines (115 loc) · 4.99 KB

infrastructure.md

File metadata and controls

145 lines (115 loc) · 4.99 KB

SocksTank Infrastructure

🇷🇺 Русская версия

Hosts

rpi5 (primary)

  • Purpose: robot tank (Raspberry Pi 5 Model B)
  • OS: Debian 13 (trixie) 64-bit, aarch64
  • CPU: BCM2712 (Cortex-A76), 4 cores, 2.4 GHz
  • RAM: 8 GB
  • Python: 3.13.5
  • Connection: WiFi (DHCP, hostname rpi5)
  • Power: XL6019E1 buck-boost (2x18650 → 5.2V → GPIO)

Key packages:

  • ultralytics, torch 2.10.0+cpu, torchvision
  • ncnn 1.0.20260114, onnxruntime 1.24.2
  • picamera2, opencv-python-headless
  • gpiozero, rpi_hardware_pwm, rpi-ws281x

Autostart: pigpiod is not required on RPi 5 (built-in GPIO).

Detailed setup: rpi5.md

rpi4 (legacy)

  • Purpose: robot tank (Raspberry Pi 4B)
  • OS: Debian 12 (bookworm), aarch64
  • Connection: WiFi

Key packages:

  • ultralytics, torch, torchvision
  • picamera2, opencv-python
  • gpiozero, pigpio, rpi_hardware_pwm
  • rpi-ws281x (LED), rpi-lgpio

File structure:

~/Freenove_Tank_Robot_Kit_for_Raspberry_Pi/
└── Code/Server/              # Main robot code
    ├── main.py                # Freenove server (TCP, client control)
    ├── car.py                 # Tank control (motors + servos + sensors)
    ├── motor.py               # Motors (gpiozero, GPIO 23/24, 5/6)
    ├── servo.py               # Servos (pigpio/gpiozero/hardware PWM, GPIO 7/8/25)
    ├── camera.py              # Camera class (preview, stream, video)
    ├── camera_detect.py       # Sock detection (moved to legacy/)
    ├── camera_shot.py         # Burst capture (moved to legacy/)
    ├── best.pt                # Trained model (copied to robot)
    ├── ultrasonic.py          # Ultrasonic sensor (GPIO 27/22, gpiozero)
    ├── infrared.py            # IR line sensors (GPIO 16/20|26/21)
    ├── led.py                 # LED control
    ├── server.py              # TCP server
    ├── tcp_server.py          # TCP server (low-level)
    ├── parameter.py           # Parameter manager (PCB/RPi version)
    └── params.json            # Configuration parameters

Autostart: start.sh launches pigpiod (GPIO daemon). Freenove server launch is commented out.

blackops (GPU server for training)

  • Purpose: YOLO model training on GPU
  • GPU: NVIDIA RTX 4070 SUPER

Working directory:

~/work/
├── test20250807_yolov8/        # Main working directory
│   ├── venv/                   # Python venv (ultralytics + torch)
│   ├── data.yaml               # Dataset config v2 (961 images)
│   ├── train.py                # Training script (100 epochs)
│   ├── bench.py                # Model benchmark
│   ├── train/valid/test/       # Dataset (train/valid/test splits)
│   ├── yolov8n.pt              # Base model YOLOv8n
│   ├── yolo11n.pt              # Base model YOLOv11n
│   └── runs/detect/            # Training results
│       ├── train/weights/best.pt   # YOLOv8n: mAP50=0.995, mAP50-95=0.885
│       └── train2/weights/best.pt  # YOLOv11n: mAP50=0.995, mAP50-95=0.96
└── ultralytics/                # Ultralytics repo clone (for development)

Networking

All devices are on the same local network. It's recommended to configure SSH aliases in ~/.ssh/config:

Host rpi5
    HostName rpi5
    User user

Host rpi4
    HostName 192.168.x.x
    User user

Host blackops
    HostName 192.168.x.x
    User user
    IdentityFile ~/.ssh/blackops

If mDNS or local DNS does not resolve rpi5 / blackops reliably, add static aliases in /etc/hosts on your development machine:

192.168.0.158 rpi5
192.168.0.124 blackops

Then the hostnames used throughout this documentation will work consistently for ssh, rsync, the web panel URL, and deploy commands.

Actual IP addresses, usernames, and SSH keys are in docs/credentials.md (not committed).

Robot hardware

GPIO pinout

Component GPIO pins Angles
Left motor GPIO 23 (forward), GPIO 24 (backward)
Right motor GPIO 5 (forward), GPIO 6 (backward)
Servo 0 (claw) GPIO 7 90° closed, 150° open
Servo 1 (lift) GPIO 8 150° up, 90° down
Servo 2 GPIO 25
Ultrasonic trigger GPIO 27
Ultrasonic echo GPIO 22
IR sensor 1 GPIO 16
IR sensor 2 GPIO 20 (PCB v1) / GPIO 26 (PCB v2)
IR sensor 3 GPIO 21

Operating modes (car.py)

  • mode_ultrasonic: Obstacle avoidance (< 45 cm — reverse + turn, otherwise forward)
  • mode_infrared: Line following (IR sensors) + object grab with claw (5-12 cm)
  • mode_clamp: Claw control (up/down/stop) via servos

Camera

  • Model: ov5647 (OmniVision)
  • Library: picamera2 + libcamera
  • Modes: preview (1280x720), still (1920x1080), stream (400x300)
  • Transform: vflip=True (camera is mounted upside down)
← Previous README Next →
SocksTank Back to README RPi 4 Setup (legacy)