Implementation of Federated Reinforcement Learning for traffic control using SUMO and a real-world testbed.
The goal of this project is to provide a platform for simple single- and multi-agent reinforcement learning (RL) and federated reinforcement learning (FedRL) for smart traffic light control. The traffic simulator we use is SUMO.
We consider a multi-agent approach for FedRL using the OpenAI Gym interface. The gym for this scenario is defined as follows:
class MultiAgentEnv(gym.Env):
def step(self, action_n: List[Any]) -> Tuple:
obs_n = list()
reward_n = list()
done_n = list()
info_n = {"n": []}
# ...
return obs_n, reward_n, done_n, info_n
For the multi-agent environment, we will use the MultiAgentEnv class provided by the RlLib API.
https://docs.ray.io/en/master/rllib-env.html#multi-agent-and-hierarchical