Don's UGV Beast

My ROS2 Jazzy-based autonomous robot built on the Waveshare UGV Beast platform with custom modular packages.

🤖 Features

• Voice-Activated Search Assistant: Say "Hey Beast" to activate, ask any question, and get spoken answers from web search
• Dual Microphone Support: Camera mic (primary) and sound board mic
• Hybrid Architecture: Local wake word detection + cloud-based speech recognition and search
• Safety Features: LiDAR-based obstacle detection with audio warnings
• Full ROS2 Integration: ESP32 bridge, odometry, headlight control, battery monitoring

🛠️ Hardware

• Platform: Waveshare UGV Beast (tank-tracked chassis)
• Computer: Raspberry Pi 5
• Sensors:
  • D500 LiDAR
  • USB Camera with microphone
  • USB PnP Audio Device (sound board)
  • IMU (on ESP32)
  • INA219 Battery Monitor
• Actuators:
  • 2x DC motors with encoders (track drive)
  • Headlights (PWM controllable)
• Controller: ESP32 for low-level motor control and sensor interfacing

📦 ROS2 Packages

beast_description

Robot URDF model and visualization

beast_bringup

Launch files and ESP32 serial bridge

• ESP32 communication via UART
• IMU data publishing
• Raw odometry publishing
• Voltage monitoring
• Headlight control service

beast_msgs

Custom messages and services

• SetLEDBrightness.srv - Headlight control

beast_motion

Motion and odometry nodes

• odom_publisher - Converts raw encoder data to odometry

beast_utils

Utility nodes

• safety_stop - LiDAR-based obstacle detection with espeak warnings

beast_interaction

Voice interaction system

• microphone_test - Test microphone functionality
• voice_assistant - Voice-activated search assistant

🎤 Voice Assistant Usage

The voice assistant uses a hybrid approach:

• Offline: Wake word detection and text-to-speech (espeak)
• Online: Speech recognition (Google API) and web search (DuckDuckGo)

How to Use:

1. Run the voice assistant node
2. Say "Hey Beast" (or "Hey Based" - it's flexible!)
3. Wait for "Yes?" response
4. Ask your question
5. Listen to the answer

Example Interaction:

You: "Hey Beast"
Beast: "Yes?"
You: "Who is the president?"
Beast: "Donald John Trump is the 47th president of the United States."

🚀 Getting Started

Prerequisites

# ROS2 Jazzy
sudo apt install ros-jazzy-ros-base

# Python dependencies
pip install SpeechRecognition ddgs --break-system-packages

# System dependencies
sudo apt install espeak sox flac python3-pyaudio

Build Instructions

cd ~/beast_ws
colcon build
source install/setup.bash

Running the Voice Assistant

ros2 run beast_interaction voice_assistant

🔧 Robot Parameters

Key measurements:

• Track separation: 0.143 m (center-to-center)
• Drive sprocket diameter: 0.0445 m
• Audio sample rate: 48kHz (resampled to 16kHz for processing)

📝 Configuration

Parameters are stored in beast_params.yaml:

• Serial communication settings
• Physical robot dimensions
• Speed limits
• Battery monitoring thresholds
• Odometry settings

🔊 Audio Configuration

The Beast uses the camera microphone (hw:0,0) for voice input:

• Supported rates: 22050-96000 Hz
• Recording rate: 48000 Hz
• Processing rate: 16000 Hz (for speech recognition)
• Channels: Mono
• Format: 16-bit PCM

🛡️ Safety Features

The safety_stop node monitors LiDAR data:

• Danger distance: 0.3m (configurable)
• Action: Publishes safety_stop Bool topic
• Audio warning: Uses espeak to announce "Warning! Obstacle detected. Stopping."

📚 Dependencies

ROS2 Packages

• rclpy
• geometry_msgs
• sensor_msgs
• nav_msgs
• std_msgs
• tf2_ros

Python Libraries

• SpeechRecognition
• ddgs (DuckDuckGo search)
• pyaudio
• subprocess (built-in)

System Tools

• arecord/aplay (ALSA)
• sox (audio resampling)
• espeak (text-to-speech)
• flac (audio encoding for Google API)

🤝 Contributing

This is a personal learning project, but suggestions and improvements are welcome!

📄 License

Apache-2.0

👨‍💻 Author

Donald Williamson (dwilliestyle@gmail.com)

🙏 Acknowledgments

• Waveshare for the UGV Beast platform
• Google Speech API for accurate transcription
• DuckDuckGo for search functionality
• ROS2 community for excellent documentation

---