Double Deep Q-Learning and SAC Based Hybrid Beamforming for 5G and Beyond Millimeter-Wave Systems

Author(s):  
Youness Arjoune ◽  
Saleh Faruque
Author(s):  
George C. Alexandropoulos ◽  
Ioanna Vinieratou ◽  
Mattia Rebato ◽  
Luca Rose ◽  
Michele Zorzi

2021 ◽  
Vol 2021 ◽  
pp. 1-10
Author(s):  
Shengli Yan

With the rapid development of information technology, facing the problems and new challenges brought by mobile Internet and Internet of things technology, as one of the key technologies of 5G, millimeter-wave mobile communication (28/38/60/70 GHz) which can realize gigabit (GB/s, or even higher) data transmission rate has also attracted extensive attention of wireless researchers all over the world, it has quickly become a research hotspot in the field of wireless communication. In the millimeter-wave massive MIMO downlink wireless sensor system, a block diagonal beamforming algorithm based on the approximate inverse of Neumann series is improved to obtain complete digital beamforming. Then, when designing hybrid beamforming, channel estimation and high-dimensional singular value decomposition are required for traditional analog and digital hybrid beamforming. A low complexity hybrid beamforming scheme is designed. An improved gradient projection algorithm is proposed in the design of analog beamforming, which can solve the problem of high computational complexity and less damage to guarantee and rate. Simulation results show that the hybrid beam terminal of the sensor reduces the number of RF links required for full digital beamforming and is as close to the spectral efficiency performance of full digital beamforming as possible. The results show that the performance of the designed hybrid beamforming scheme can still be close to that of the pure digital beamforming scheme without involving channel estimation and SVD decomposition.


Sign in / Sign up

Export Citation Format

Share Document