How mcgill-robotics/auv-2025 works
Based on our analysis of the file tree and README, this project is an autonomous underwater vehicle (AUV) software system built for the Douglas AUV intended for RoboSub2025. It is built on top of ROS (using a Catkin workspace) and is structured as a modular robotics software suite with several core functionalities (controls, planner, propulsion, sensors, state estimation, vision, simulation) combined with Docker‐based deployment support for different target platforms (e.g., Jetson, poolside, simulated environments, etc.).
And Group the Main Components
• Hardware & Deployment Containers:
• Docker directory shows multiple Dockerfiles (for different hardware targets like Jetson, poolside, trottier, sim). Represent these as separate “deployment containers” that encapsulate the software runtime environments.
• ROS-Based Software Modules (inside catkin_ws):
• Controls: Contains PID controllers, quaternion handling, and server components (under catkin_ws/src/controls).
• Planner: Handles mission planning, joystick input, and various search/maneuver strategies.
• Propulsion: Deals with thrust mapping and driving the physical propulsion mechanisms.
• Sensors: Manages sensor data acquisition (IMU, DVL, hydrophones, etc.) and calibration.
• State Estimation: Takes in sensor input to determine the vehicle’s state (localization, depth, orientation).
• Vision: Handles camera input, object detection (YOLO models are referenced), image processing, lane marker measurements.
• Simulation: Provides a simulation environment using custom packages (sim folder) and an endpoint for TCP communication with Unity (ROS TCP endpoint modules).
• Communication Infrastructure:
• ROS Launch files: Launch files in each module indicate orchestration of node startup sequences.
• Supporting Infrastructure:
• Scripts: Helper scripts exist for tasks like setting environment variables, clearing ports, etc.
• CI/CD Workflow: GitHub workflows and associated files show automated deployment/CI processes.
• Configuration Files: YAML configuration files exist for sensors and deployment, emphasizing adjustable parameters.
Detail Relationships and Data Flows
• For example, sensor nodes (in Sensors) send data to State Estimation and Planner.
• Planner commands through Controls which subsequently drive Propulsion.
• Vision components perform object detection and feed results to the Planner or even State Estimation.
• Docker containers encapsulate these modules, so highlight interactions between containerized environments and the hardware deployment targets (e.g., Jetson board handling real sensor input).
• Interactions between simulation and real-time control: The Sim package provides a TCP endpoint which could be integrated with simulation environments (e.g., Unity) for testing.
Architectural Patterns and Design Principles
• Modular Design: The clear separation of concerns (controls, sensors, planning, etc.) follows a modular design pattern typical in ROS-based projects.
• Message-Driven Communication: Emphasize the usage of ROS messaging patterns (topics, actions, and messages defined in auv_msgs).
• Containerization and Deployment: Docker’s presence under multiple folders reflects the need for reproducible environments across different platforms (embedded, simulation, etc.).
Specific Technologies and Frameworks
• ROS (Robot Operating System) and Catkin: The backbone for node management and inter-module communication.
• Docker: Used for containerized builds and ensuring consistency across platforms.
• Specific Sensor Drivers & Models: YAML config files for sensors, deep learning models for vision (YOLO).
• CI/CD Workflows: Provided in the .github and Docker directories to support automated testing and deployment.