How fontainebleau2021/agv_2023 works
Project type and purpose
• This is a ROS‐based AGV (automated guided vehicle) integration framework (agv_2023_pkg), organized as a catkin workspace with multiple ROS packages.
• It brings together sensor drivers (sensor_pkg), SLAM/localization (SLAM_pkg), mapping (map_pkg), navigation planning (navigation_pkg), control and actuation (control_pkg + motor_control), network/cloud connectivity (networksocket_pkg), orchestration services (service_pkg), and shell scripts to launch sensors (sensor_run_sh).
High-level package breakdown
• sensor_pkg: IMU, GPS, LiDAR (Livox/Ouster), camera (ZED, USB), Marvelmind, NMEA driver
• SLAM_pkg: LIO‐SAM SLAM algorithm (with relocalization)
• map_pkg: static map loaders & pointcloud publishers for RViz
• navigation_pkg: costmaps, global planner, TEB local planner, “navi_master” orchestration
• control_pkg: UGV SDK (bunker, hunter, scout, tracer demos), ROS control nodes, motor_control for servos
• networksocket_pkg: ROS socket clients and web UI (robot_html) for remote visualization/control
• service_pkg: RPC servers (agr_service) exposing services for starting/stopping nodes, SLAM service
• sensor_run_sh: convenience shell scripts to bring up each sensor
Key architectural patterns
• ROS pub/sub: sensors → topic streams → SLAM/localization → odometry (/tf), pointclouds, images → navigation nodes → velocity commands → control nodes → hardware/SDK
• ROS services/actions: agr_service, save_map service, network socket RPC
• Parameterization & launch: each package has dedicated launch/ folder, YAML parameter files under config/
• Modularity: each functional domain in its own package; common messages under msg/, service definitions under srv/
Main components
A. ROS Master (central graph manager)
B. Sensor Layer (sensor_pkg + sensor_run_sh)
• IMU node(s) (openzen_sensor, imu_integrator)
• GPS node(s) (nmea_navsat_driver, imu_gps_localization)
• LiDAR nodes (livox_ros_driver, ouster_ros)
• Camera nodes (zed-open-capture, usb_cam)
• Marvelmind beacon node
C. SLAM/Localization Layer (SLAM_pkg/lio_sam_devel)
• Input: /imu/data, /velodyne_points, /livox/lidar
• Output: optimized pose (/aft_mapped_to_init), cloud_info topic
D. Mapping & Visualization Layer (map_pkg)
• map_load publishes static occupancy map & RViz config
• pcd_to_rviz and publish_pointcloud nodes publish demo point clouds
E. Navigation & Planning Layer (navigation_pkg)
• Costmap generator → local & global planners → TEB planner
• navi_master orchestrates state machine, subscribes odometry, publishes /cmd_vel or /ackermann_drive
F. Control & Actuation Layer (control_pkg + motor_control)
• UGV SDK library wraps CAN/serial ports → bunker_base_node etc.
• motor_control node exposes TCP/ROS interface, subscribes velocity commands, drives servo motors
G. Service & Orchestration Layer (service_pkg)
• agr_service node provides ROS service /agr_service to start SLAM/navigation clients
H. Network & Web UI Layer (networksocket_pkg)
• robotsocket C++ node provides TCP/UDP socket interfaces, publishes/subscribes ROS topics remotely
• robot_html web client (JavaScript) connects to robotsocket, visualizes map & robot pose
Interactions and data flows
• Sensors publish raw data streams → SLAM & mapping nodes consume them
• SLAM outputs pose & pointcloud info → navigation nodes read pose for planning
• Navigation publishes velocity commands → control nodes translate to motor driver messages
• Service nodes allow on‐demand start/stop via ROS services
• Network socket nodes mirror selected topics to remote client; web UI subscribes to these via WebSockets
Technologies & frameworks
• ROS Melodic (nodes in C++/Python)
• LIO‐SAM (GTSAM, PCL, Eigen)
• Livox & Ouster SDKs
• ZED Open Capture (C++ + ROS API)
• Web UI: roslibjs, ros2djs, eventemitter2
• UGV SDK: C++ async ports, protocol parsers
• Dockerfile (SLAM_pkg) for reproducible container
• CMake / catkin build system; parameters in YAML; RViz configs
• Clear labels: component names, topic/service names
• External dependencies (e.g., Livox/Ouster hardware, Web browser)