The simulation runs in ns-3 (C++) and communicates with a Python RL agent via the ns3-ai interface. On each step the agent receives SINR values for every (BS, UE) pair and decides the TX power for each base station.
Python RL Agent (Ray/RLlib)
↕ ns3-ai socket
ns-3 LTE Simulation (C++)
└── PowerControlMobComEnv
├── Observation: SINR matrix [nBS × nUE]
├── Action: TX power per BS [0..46 dBm]
└── Reward: coverage + efficiency
Covered by Apache 2.0 License.