ROS2 Launch Inspection Tool

This project provides tools to record the execution of ROS 2 launch. Then, analyze and replay the launch execution.

Install Pre-built Packages (Recommended)

Find the latest pacakges in the Release page and follow the instructions.

Install from Source

Prerequisites

• Rust toolchain

  Visit rustup.rs and install rustup.

• rye

  Visit rye.astral.sh to install it. It is used to setup a Python virtual environment and to manage Python project.

• procpath

  Visit the pypi site to install this tool. This is used for resource usage profiling.

Build This Project

Clone the project Git repository.

git clone https://github.com/NEWSLabNTU/ros-launch-perf.git

Compile the project and install the dump_launch and play_launch commands on the system.

make install

You can uninstall these commands afterwards.

make uninstall

Usage

The workflow goes in the dump step and then the play step. The dump step cans the launch file and creates a record.json. The play step reads the dump and executes the launch plan. The dump only needs to be done once.

Step 1: Record a Launch Execution

The dump step runs in the same way we run a launch file. In this Autoware simulation example, simply replace ros2 launch with dump_launch.

## This is the original launch command.
# ros2 launch \
#     autoware_launch planning_simulator.launch.xml \
#     map_path:=$HOME/autoware_map/sample-map-planning \
#     vehicle_model:=sample_vehicle sensor_model:=sample_sensor_kit

dump_launch \
    autoware_launch planning_simulator.launch.xml \
    map_path:=$HOME/autoware_map/sample-map-planning \
    vehicle_model:=sample_vehicle sensor_model:=sample_sensor_kit

Step 3: Play the Launch Record

This command loads the record.json and perform the launch execution.

play_launch

Profiling Resource Usage

Start the launch first by just play. Then, run this command to profile per-process resource usage into profiling.sqlite database file. Press Ctrl-C to terminate the profiling.

make profile

Afterwards, you can generate SVG timecharts using this command.

make plot

License

This software is distributed under MIT license. You can read the license file.