Systemuebersicht¶

Der AMR besteht aus einem Raspberry Pi 5 als zentralem ROS2-Rechner und zwei ESP32-S3-Knoten fuer echtzeitnahe I/O- und Regelaufgaben.

Drei-Ebenen-Modell¶

Ebene A arbeitet autonom — der CAN-Notstopp funktioniert ohne Pi 5. Ebene B setzt auf micro-ROS-Kommunikation mit A auf. Ebene C ist vollstaendig optional und ueber Launch-Argumente zuschaltbar.

Hauptkomponenten¶

• Pi 5: ROS2, SLAM, Navigation, Dashboard-Bridge, optionale Vision, Audio-Feedback und TTS-Sprachausgabe (MAX98357A I2S)
• Drive-Knoten: Motoransteuerung, Encoder, PID, Odometrie, LED
• Sensor-Knoten: Ultraschall, Cliff, IMU, Batterie, Servo (PCA9685 PWM)
• Peripherie: LiDAR, Kamera, optionale Hailo-/Gemini-Pipeline

Systemdiagramm¶

graph TB
  subgraph C["Ebene C — Intelligente Interaktion (optional)"]
    HAILO["Host: Hailo-8L Runner<br>YOLOv8 @ 5 Hz"]
    UDP["hailo_udp_receiver"]
    GEMINI["gemini_semantic_node<br>Gemini 2.0 flash-lite"]
    TTS["tts_speak_node<br>gTTS → MAX98357A"]
    RESPEAKER["ReSpeaker DoA"]
    HAILO -->|UDP:5005| UDP
    UDP -->|"/vision/detections"| GEMINI
    GEMINI -->|"/vision/semantics"| TTS
  end

  subgraph B["Ebene B — Bedien- und Leitstandsebene"]
    subgraph DOCKER["Docker-Container (ros:humble)"]
      AGENT_D["micro-ROS Agent<br>/dev/amr_drive"]
      AGENT_S["micro-ROS Agent<br>/dev/amr_sensor"]
      SLAM["SLAM Toolbox"]
      NAV["Nav2"]
      ODOM_TF["odom_to_tf"]
      CLIFF["cliff_safety_node"]
      AUDIO["audio_feedback_node"]
      CAN_BR["can_bridge_node"]
      DASH_BR["dashboard_bridge<br>WSS:9090 / MJPEG:8082"]
    end
    BROWSER["Benutzeroberflaeche<br>React/Vite HTTPS:5173"]
    BROWSER <-->|WSS:9090| DASH_BR
  end

  subgraph A["Ebene A — Fahrkern"]
    DRIVE["Drive-Knoten ESP32-S3<br>Core 0: micro-ROS<br>Core 1: PID 50 Hz"]
    SENSOR["Sensor-Knoten ESP32-S3<br>Core 0: micro-ROS<br>Core 1: Sensorerfassung"]
    CAN_BUS["CAN-Bus 1 Mbit/s<br>Cliff 0x120 / Battery 0x141"]
    DRIVE <--> CAN_BUS
    SENSOR <--> CAN_BUS
  end

  RESPEAKER -->|"/sound_direction"| AUDIO
  DRIVE <-->|"UART 921600"| AGENT_D
  SENSOR <-->|"UART 921600"| AGENT_S

  style C fill:#1a0a2e,stroke:#00E5FF,color:#cdd9e5
  style B fill:#111D2B,stroke:#00E5FF,color:#cdd9e5
  style A fill:#0B131E,stroke:#00E5FF,color:#cdd9e5
  style DOCKER fill:#162435,stroke:#517C96,color:#cdd9e5

Docker-Container-Grenzen¶

graph LR
  subgraph CONTAINER["Docker-Container (arm64)"]
    direction TB
    NODES["ROS2 Nodes:<br>micro-ROS Agents, SLAM, Nav2,<br>cliff_safety, dashboard_bridge,<br>audio, can_bridge, vision, TTS"]
    DEV["Devices:<br>/dev/amr_drive, /dev/amr_sensor,<br>/dev/ttyUSB0, /dev/snd"]
    VOL["Volumes:<br>my_bot (rw), scripts (ro),<br>dashboard (ro), Named Volumes"]
    NET["Network: host, Privileged: true"]
  end

  subgraph HOST["Host (Pi 5, Debian Trixie)"]
    HAILO_H["host_hailo_runner.py<br>Python 3.13, Hailo-8L"]
    VITE["Dashboard Dev-Server<br>npm run dev :5173"]
    CERT["mkcert-Zertifikate"]
  end

  HOST -->|localhost| CONTAINER
  CONTAINER -->|localhost| HOST

  style CONTAINER fill:#111D2B,stroke:#00E5FF,color:#cdd9e5
  style HOST fill:#0B131E,stroke:#517C96,color:#cdd9e5

TF-Baum¶

graph LR
  ODOM["odom"] -->|"dynamisch, 20 Hz<br>odom_to_tf"| BASE["base_link"]
  BASE -->|"statisch<br>x=0.10, z=0.235, yaw=pi"| LASER["laser"]
  BASE -->|"statisch<br>x=0.10, z=0.08"| CAM["camera_link"]
  BASE -->|"statisch<br>x=0.15, z=0.05"| US["ultrasonic_link"]

  style ODOM fill:#111D2B,stroke:#00E5FF,color:#00E5FF
  style BASE fill:#111D2B,stroke:#00E5FF,color:#00E5FF
  style LASER fill:#111D2B,stroke:#00FF66,color:#00FF66
  style CAM fill:#111D2B,stroke:#517C96,color:#517C96
  style US fill:#111D2B,stroke:#517C96,color:#517C96

Der LiDAR ist 180 Grad gedreht montiert — die TF-Transformation (yaw=pi) kompensiert dies.