AI for cell shaping in mobile networks

Cell shaping is used to configure radio antenna parameters (in this case, electrical downtilt) to improve the service quality.

Julien Forgeat explores an RL agent that automates the cell shaping. More precisely, the agent aims at tuning the tilt of multiple antennas to be generalized and adaptive to the random traffic demand and hotspots in an urban map. The training has been accelerated by leveraging RLlib’s scalability, an open source distributed RL library driven by UC Berkeley’s RISELab. RLlib’s support for distributed execution is critical, since this radio simulator is very compute intensive. Julien demonstrates that the RL agent’s tuning is well adapted to any scenario and outperforms the baseline algorithms.