Example

We provide several experiment setups in the examples folder. However, here we want to walk you through the most basic example without a planner that has a dependency on other packages.

Navigate to example

First, we must ensure that everything is installed and enter the virtual environment.

cd /path/to/localPlannerBench
poetry install
poetry shell
cd experiments/example

Nice! In this folder, you should see a setup directory and a results directory. The first contains the configuration files for the planner planner.yaml and the experiment exp.yaml. Go ahead and look a bit for yourself where the obstacles, goals and parameters to the planner are specified.

Run the example

Once you have familiarized yourself with the directories and files, you can run the example. For that execute

runner -c setup/exp.yaml -p setup/<planner_name>.yaml --render

When you run the above line, the experiment starts with the configuration files you have provided through the command line interface, namely -c setup/exp.yaml specifies the experiment and -p setup/<planner_name>.yaml specifies the planner. The flag --render indicates that the experiment should be rendered to your screen. Once the experiment has finished or you have stopped it prematurely, a new folder is created in the results directory. In there, you will find all necessary information on what exactly happened during the expereiment and all the configuration you have set.

Postprocessing

In the previous step, you have run the experiment and you now want to access the performance of the particular planner. This is what the postProcessor is all about. The postprocessor uses the results folder that was created after running the experiment to evaluate user-specified metrics. Familiarize yourself a bit with the content of the folder you want to postprocess. When you are ready, you can simple invoke the postProcessor by

post_process --exp results/<name_experiment> -k time2Goal pathLength --plot

In the above line, the argument --exp results/<name_experiment> tells the postProcessor which experiment to process. Key-performance-indicators are listed behind -k option. The flag --plot that a plot of the trajectory should be created. Navigate to the experiment folder in your explorer to access the plots and evaluations. The evaluations are stored in postProcess.yaml.

Conclusion

Nice! You have run your first experiment for testing a local motion planning algorithm. Feel free to add your own method or additional functionility.