Core Concepts of ROS (Continued)

TF2 is the ROS coordinate frame transform library. It maintains a tree of coordinate transformations over time, allowing any part of the system to query the 3D pose of any frame relative to any other frame at any past timestamp.

Every physical component of the robot has a named frame: base_link (robot body), base_footprint (projection onto ground), camera_link, laser_link, odom, map, end_effector. The TF tree connects these with static or dynamic transforms.

Broadcasting transforms is done via tf2_ros.TransformBroadcaster (dynamic, e.g. wheel odometry updates odom→base_link) or tf2_ros.StaticTransformBroadcaster (static, e.g. camera mounting position). The robot_state_publisher node reads the URDF and publishes transforms for all joints automatically.

Querying transforms:

ros2 run tf2_ros tf2_echo base_link camera_link     # print live transform
ros2 run tf2_tools view_frames                       # generate TF tree PDF

TF2 is critical for: fusing LIDAR scans (expressed in laser_link) with odometry (expressed in odom) — you transform both into map frame before combining. For manipulation, TF2 converts a detected object pose (camera frame) into the robot base frame for grasping. Without TF2, every node would need to maintain its own transform bookkeeping — a source of many bugs.

Launch files in ROS 2 are Python scripts (not XML as in ROS 1) that programmatically describe an entire system configuration: which nodes to start, with what parameters, in what namespaces, with what remappings, and in what order.

Key building blocks:

• LaunchDescription — the root container returned by the generate_launch_description() function
• Node — declares a node to launch: package, executable, name, namespace, parameters, remappings
• IncludeLaunchDescription — nests another launch file, enabling modular composition
• DeclareLaunchArgument — declares a command-line argument (e.g. use_sim_time:=true)
• LaunchConfiguration — reads the value of a declared argument at runtime
• GroupAction with PushRosNamespace — applies a namespace to a group of nodes

from launch import LaunchDescription
from launch_ros.actions import Node

def generate_launch_description():
    return LaunchDescription([
        Node(
            package='my_robot_bringup',
            executable='drive_controller',
            name='drive_node',
            parameters=[{'max_speed': 1.5, 'use_sim_time': False}],
            remappings=[('/cmd_vel', '/robot/cmd_vel')]
        ),
    ])

Launch files replace shell scripts for starting robot systems, providing dependency ordering, conditional logic, and argument passing in a maintainable Python format.

ROS 2 ships with a rich set of command-line and GUI tools for development, debugging, and data management:

Data logging and replay: ros2 bag record -a records all active topics to an MCAP file. ros2 bag play replays them, allowing offline algorithm development without a physical robot. Bags can be filtered, merged, and converted with ros2 bag convert.

3D Visualisation: RViz2 is the primary 3D visualiser. It displays robot models (URDF), sensor data (LIDAR scans, camera images, point clouds), transforms (TF tree), navigation costmaps, and planned paths in real time. Panels and displays are configurable and saveable as .rviz config files.

rqt Plugin Framework: rqt provides a Qt-based plugin system with dozens of plugins: rqt_graph (computation graph), rqt_plot (live time-series plots of any numeric topic), rqt_console (log message viewer with severity filtering), rqt_image_view (camera feed viewer), rqt_reconfigure (live parameter editing).

Diagnostics: ros2 doctor analyses the ROS 2 installation, environment variables, DDS settings, and running system for common configuration problems. ros2 interface show prints the full definition of any message, service, or action type.