The intention of reinforcement learning is to master a policy, and that is a mapping from states to actions, that maximizes the predicted cumulative reward as time passes. Metropolitan areas already make large quantities of data (from security cameras and environmental sensors) and currently cont