How watonomous/wato_monorepo works
Project type and purpose
• This is a Dockerized ROS2 “monorepo” for an autonomous vehicle software pipeline.
• It bundles multiple ROS2 packages (perception, world modeling, planning, control, simulation, interfacing, sample apps) into Docker images and composes them via watod (a CLI wrapper around Docker Compose).
File structure
• Top-level orchestration:
• watod, watod-setup-env.sh: CLI & env management
• modules/*.yaml: per-module Docker Compose fragments
• docker/...: Dockerfiles for each ROS2 service
• Source code:
• src/action, src/perception, src/world_modeling, src/interfacing, src/simulation, src/samples, src/wato_msgs
• Each is a ROS2 package exposing nodes, topics, services.
• Docs and CI:
• docs/: developer & setup guides
• .github/: CI workflows, linters
README
• Explains “modules” which map 1:1 to functional domains (perception, planning, etc.).
• Describes watod up/run as the single command to spin up containers.
• Emphasizes environment selection via ACTIVE_MODULES → .env → Docker Compose.
Main components
A. CLI & Orchestration Layer
• watod CLI
• .env generator (watod-setup-env.sh)
• Docker Compose module loader (selects modules/*.yaml)
B. Container Layer
• Action (behaviour_planning, global_planning, local_planning, model_predictive_control)
• Interfacing (CAN, sensor)
• Perception (camera_detection, depth_estimation, lane_detection, lidar/det, radar_det, semseg, tracking, traffic_light)
• World Modeling (hd_map, localization, occupancy, occupancy_segmentation, motion_forecasting)
• Simulation (Carla bridge, sample node, viz, notebooks)
• Infrastructure (data_stream, foxglove, VNC)
• Samples (cpp_*/py_*)
• Each container runs one or more ROS2 nodes, mounts the corresponding src/... folder.
C. Communication Layer (ROS2 DDS network)
• Shared pub/sub bus: topics and services built on wato_msgs.
• External bridges: e.g., CARLA simulator <-> ROS2, Foxglove bridge.
D. External systems & tools
• CARLA simulator
• Foxglove visualization
• Host (for VNC, port forwarding)
Relationships and interactions
• All containers join a virtual network (show a “ROS2 network” backbone).
• Sensors → Interfacing → Perception → World Modeling → Action → Model Predictive Control → Actuator commands
• Simulation ↔ CARLA bridge feeds simulated sensors and consumes control commands.
• That messages are typed by wato_msgs packages.
Architectural patterns and principles
• Microservices: each ROS2 node runs in its own container; modules selected on demand.
• Pub/Sub: decoupled communication via ROS2 topics.
• Extensibility: new modules equate to new Docker-Compose fragments + Dockerfiles + ROS2 packages.
• Monorepo: source mounts drive live code changes in containers without rebuilding images.
Technologies & frameworks
• ROS2 (rclcpp, rclpy)
• Docker & Docker Compose
• Python & C++ (various ROS2 packages)
• CARLA simulator, Foxglove
• Tensorrt for inference in perception modules
• CI: GitHub Actions