This is an OpenAI Gym environment following the specififactions of the problem defined in the Dynflor paper. The purpose of this environment is to prove that a reinforcement learning approach for Multi Robot Patrolling is highly feasible and further research conducted in this area could lead to huge improvements.
The environment can be used as every other gym environment and we encourage you to further develop it and apply your own reinforcement learning approaches. All the necessary observations about the action and state space can be found here.