Autoware-Autotune

alpha release

Installation

1. Set up aichallenge repo with the following reference.

https://github.com/soyaoki/aichallenge2023-sim/blob/main/documentation/docfx_project/setup/index.md

2. Copy AWSIM-related executables such as AWSIM.x86_64 under /aichallenge2023-sim/docker/aichallenge/aichallenge_ws/AWSIM/.

3. Clone this repository.

cd ~/aichallenge2023-sim/docker/aichallenge/aichallenge_ws/
git clone https://github.com/soyaoki/Autoware-Autotune

4. Run the docker container.

cd ~/aichallenge2023-sim/docker
sudo bash run_container.sh

5. Intstall requirements.

pip install optuna optuna-dashboard

6. Edit target parameters in target_param.yaml.

7. Run param_autotuner.py.

cd /aichallenge/aichallenge_ws/Autoware-Autotune/
python3 param_autotuner.py 

(Optional) 8. Run dashboad on anothoer terminal.

cd ./aichallenge2023-sim/docker
sudo bash run_container.sh
pip install optuna optuna-dashboard
cd /aichallenge/aichallenge_ws/Autoware-Autotune/
optuna-dashboard sqlite:///Autoware_turning_study.db

Then access "http://127.0.0.1:8080/". You can see the process of parameter optimization on Optuna dashboard.

NOTE

・The default iteration is 50 iterations.　（TODO: Make it possible to pass the number of iterations as an argument.）

study.optimize(lambda trial: objective(trial, param_ranges, param_files, target_param), n_trials=50)

・Comments in Autoware parameter files (e.g., obstacle_avoidance_planner.param.yaml) will disappear.

・一部治安が悪い

# シャットダウン
# 現在のPythonプロセス以外のすべてのプロセスをキル
current_process = psutil.Process()
for process in psutil.process_iter(attrs=['pid', 'name']):
    if process.info['pid'] != current_process.pid:
        process.terminate()  # プロセスを終了させる