This is an environment for use with Veins-Gym. The scenario uses DCC, the aim is to learn optimal parameters. For this, the observations are the average channel busy time, reward is a metric derived from the vehicles' age of information, and the action are the CBR values used by DCC for its transition.
Initial steps:
- install dependencies of the simulation: SUMO (v1.6.0) and OMNeT++ (v5.6.*), such that you can run Veins (v5.1, bundled with veins-gym)
- install the dependencies listed in
requirements.txt - build the simulation:
snakemake -jall - and run the example:
agents/trivial.py.
For a deeper look into the simulation, see its configuration (scenario), and the GymConnection class.
Check out veins-gym, which serves as a foundation for this work. The veins-gym repository also contains a Dockerfile that can be used to build a containerized environment to run the DCC-Env in.