How ucla-robot-intelligence-lab/realman-env works
Project type and purpose
• This is a robotics middleware workspace built on ROS 2 Foxy. It combines Python and C++ packages to control AGVs, dual-arm robotic manipulators, Realsense cameras, servo drivers, and demos (both real and simulated via Gazebo & MoveIt). It is not a web app but a multi‐package ROS 2 system orchestrating sensors, controllers, drivers, planners, and simulators.
Top-level components
a. ROS 2 Core
• The ROS 2 middleware (DDS) that underpins all inter‐node communication.
b. Sensor subsystem
• Packages: ros2_realsense2 → realsense2_camera (node publishes Image/PointCloud/IMU topics)
c. Controller subsystem
• python_controller (keyboard_base_controller, joystick_base_controller) publishes control topics
d. AGV & mobility subsystem
• ros2_agv_robot (C++ nodes: exec_pre_task, exectask, goto_mark, monitor, stepctrl)
e. Dual-arm manipulator subsystem
• ros2_rm_robot packages:
• dual_rm_driver (C++ rm_driver node providing low-level servo interface)
• dual_rm_control (trajectory generation via cubicSpline)
• dual_rm_moveit_config + demos (MoveIt planners & demo nodes)
• dual_rm_description + dual_rm_gazebo (URDF/XACRO description and Gazebo plugins)
• rm_ros2_interfaces (custom message definitions for arm commands & states)
f. Servo driver subsystem
• ros2_servo_driver (C++ servo_driver_node, example clients)
g. Orchestration & demo subsystem
• ros2_total_demo (total_demo_node, scripts that tie sensors, AGV, arms together)
h. External tools
• RViz (visualization)
• Gazebo (simulation)
• MoveIt2 (motion planning)
Relationships and interactions
• python_controller subscribes user inputs and publishes control commands (e.g., /cmd_vel or arm-specific topics).
• AGV & arm drivers subscribe command topics and send actuator commands over serial/ethernet to hardware or simulated interfaces.
• dual_rm_driver exposes ROS 2 topics/services for low-level control; dual_rm_control subscribes state topics to compute smooth trajectories.
• MoveIt2 nodes subscribe the robot’s joint state, offer planning services/actions, and send joint trajectories to dual_rm_control or dual_rm_driver.
• ros2_total_demo node aggregates sensor feeds, computes object poses, then dispatches tasks to AGV controller and dual-arm planner.
• Gazebo plugin node (mimic_joint_plugin) subscribes joint commands for simulated robot.
• RViz connects to topics for visualization.
Architectural patterns and principles
• Pub/Sub messaging for decoupled nodes.
• Service/Action interfaces for planning and pre-programmed tasks.
• Layered driver → control → planning → orchestration.
• Modular ROS 2 packages; clear separation of description (URDF), driver, control, planning, and demos.
• Use of ROS 2 launch files to compose systems.
a. Sketch the ROS 2 Core in the center (a cloud symbol).
b. Around it place:
– AGV subsystem (bottom-left): ros2_agv_robot nodes → subscribe cmd_vel → drive AGV.
– Dual-arm subsystem (bottom-right): driver, control, description, moveit config, gazebo plugin.
– Orchestration & demo (center-right): ros2_total_demo node.
e. Launch files used to bring up each subsystem (e.g., start_65_agv.launch.py, demo.launch.py).