Multi-Antenna Tuning Simulation Platform by Deep Reinforcement Learning

Recently, communication technology is highly developed. The communication convenience that people enjoy is relying on a large number of base station antenna devices set up by major operating companies. When the parameters of the antennas in local area are adjusted reasonably, the Reference Signal Receiving Power (RSRP), Signal to Interference plus Noise Ratio (SINR) and other related indicators in the region will be at a reasonable level to ensure the communication quality of users. However, the number of antennas is huge, and manual adjustment of various parameters is bound to cost a lot of money and time. Therefore, a multi-antenna simulation platform is built in this paper, which applies reinforcement learning to self-learn the parameters of the antennas, and learns an optimal antenna tuning policy. Finally, the results are migrated to real antenna scenarios, which saves the cost of antenna adjustment and has high economic value. This paper proposed a method that apply multi-agent reinforcement learning technology to the multi-antenna tuning scene, and achieved good results in the simulation scene.