| bw_interop | ||
| clang_interop | ||
| latency_graph | ||
| matching | ||
| message_tree | ||
| misc | ||
| tracing_interop | ||
| .gitignore | ||
| batch_analyze.bash | ||
| make-plots.ipynb | ||
| README.md | ||
| requirements.txt | ||
| trace-analysis.ipynb | ||
MA Autoware Trace Analysis
Automatically extract data dependencies and end-to-end (E2E) latencies from ROS2 trace data and source code.
Prerequisites
- Python 3.10 or newer (this is crucial!)
- JupyterLab or Jupyter Notebook
- ROS2 Tracing
- Tracetools Analysis
python3-babeltraceandpython3-lttng, e.g. viasudo apt install, for Python 3.10 (this requires either Ubuntu 22.04 or custom installation)
Requirements for Optional Features
ros2_internal_dependency_analyzerfor finding intra-node dependencies via static analysis, which can then be utilized by this toolros2 multitopic bwfor recording message sizes which can then be processed by this tool
Installation
# Make sure you are running Python 3.10 and the corresponding pip:
python3.10 -V
pip3.10 -V
pip3.10 install -r requirements.txt
Make sure that ROS2 Tracing and Tracetools Analysis were compiled with colcon build [...] --symlink-install [...]. Without --symlink-install, the build folder structure will be incorrect and the libraries required by this tool cannot be located.
Usage
The trace_analysis.ipynb notebook is the entry point for users.
Configure the notebook according to the comments in the user settings cell.
Settings can either be changed in the notebook or via environment variables:
# In the notebook (User Settings cell near the top):
TR_PATH = "path/to/trace/dir"
E2E_ENABLED = True
...
# In the shell of your choice (choose Bash!):
# Each setting is named the same as in the notebook but prefixed by "ANA_NB_".
# For strings, quotes have to be included IN THE VARIABLE VALUE, i.e. double quotation "'...'"
# has to be used in most cases
ANA_NB_TR_PATH="'path/to/trace/dir'"
ANA_NB_E2E_ENABLED="True"
You can run the notebook via the "Run All" command in Jupyter or you can execute it headless from the command line:
jupyter nbconvert --to notebook --execute trace-analysis.ipynb
The notebook also supports invocation via Papermill.
The output files are found in the configured output dir (default: out/).
Inputs are processed and cached in cache/.