For intermittent deployment missions, this planner enables global mission planning and localized, adaptive optimization from observations. The base of the fumes planner is a trajetory optimizer, that uses black-box optimization methods to select trajectory parameters the optimize a given reward function.
The optimizer can also chain together a given trajectory type to target dynamic phenomnea.
Project Python dependencies are managed by pipenv. You may need to install pipenv on your machine. To generate your fumes virtual environment with all dependencies installed, navigate to src/planning/fumes on your local machine and run the command: pipenv install. Then, to activate your environment, run pipenv shell from the same directory. To install new dependencies to the virtual environment, run pipenv install <package_name>.
You may want to add the following to your ~/.bashrc (or appropriate file):
alias fumes="cd /<path to fumes repo>/fumes; pipenv shell"
Then, when starting a new terminal, typing fumes will go to the correct directory, activate the fumes Python virtual environment, and load the necessary environment variables, located in .env.
The .env file sets the environment variables FUMES_DATA and FUMES_OUTPUT, which is the path to data and outputs from robot simulations respectively. Data files supporting the fumes code (simulation models, etc.) should be places in fumes/data and output files from, e.g., robot simulation runs, will be saved in fumes/output.
Example robot simulations can be found as Python scripts in fumes/fumes/examples.
Test cases are found in fumes/fumes/tests and can be run with pytest, e.g. pytest fumes/tests.