A Deep Learning and Geospatial Data-Based Channel Estimation Technique for Hybrid Massive MIMO Systems

Abstract

This paper presents a novel channel estimation technique for the multi-user massive multiple-input multiple-output (MU-mMIMO) systems using angular-based hybrid precoding (AB-HP). The proposed channel estimation technique generates group-wise channel state information (CSI) of user terminal (UT) zones in the service area by deep neural networks (DNN) and fuzzy c-Means (FCM) clustering. The slow time-varying CSI between the base station (BS) and feasible UT locations in the service area is calculated from the geospatial data by offline ray tracing and a DNN-based path estimation model associated with the 1-dimensional convolutional neural network (1D-CNN) and regression tree ensembles. Then, the UT-level CSI of all feasible locations is grouped into clusters by a proposed FCM clustering. Finally, the service area is divided into a number of non-overlapping UT zones. Each UT zone is characterized by a corresponding set of clusters named as UT-group CSI, which is utilized in the analog RF beamformer design of AB-HP to reduce the required large online CSI overhead in the MU-mMIMO systems. Then, the reduced-size online CSI is employed in the baseband (BB) precoder of AB-HP. Simulations are conducted in the indoor scenario at 28 GHz and tested in an AB-HP MU-mMIMO system with a uniform rectangular array (URA) having 16 × 16 = 256 antennas and 22 RF chains. Illustrative results indicate that 91.4% online CSI can be reduced by using the proposed offline channel estimation technique as compared to the conventional online channel sounding. The proposed DNN-based path estimation technique produces same amount of UT-level CSI with runtime reduced by 65.8% as compared to the computationally expensive ray tracing. The imperfection of UT-level CSI introduced by the DNN-based path estimation technique is mitigated by the FCM clustering technique, where the AB-HP using offline UT-group CSI generated by the DNN-based channel estimation model and ray tracing based model for the RF beamformer achieves, respectively, 98.7% and 99.1% sum-rate performance of the fully digital precoding (FDP) technique using full-size online CSI.

Cluster Index Modulation for mmWave Communication Systems

“…is the trace operation. It is worth mentioning that, channel characteristics at the transmitter can be obtained via both online and offline approaches (Zheng et al, 2018;Wang et al, 2015;Zhu et al, 2021). The received signal vector after analog combiner can be stated as follows:…”

Deep Learning based Multi-User Power Allocation and Hybrid Precoding in Massive MIMO Systems

“…are the weight matrix and bias vector, respectively. In order to fit the output layer predictions between 0 and 1 as in the output labels expressed in (14), we employ the sigmoid function at the output layer (i.e., f σ (x) = 1 1+e −x [11]). Thus, the predicted power values for K downlink UEs via the DNN architecture are written as:…”

“…As a key driving force for artificial intelligence (AI), deep learning has been successfully applied in many fields including computer vision, speech recognition and natural language processing [11]. Hence, the success of deep learning also motivates its applications in wireless communication systems [12]- [14]. For instance, deep learning has been applied for signal detection [12], resource management [13], channel estimation [14].…”

Energy-Efficient Throughput Maximization in mmWave MU-Massive-MIMO-OFDM: Genetic Algorithm based Resource Allocation

“…Based on the URA structure, Mx and My denote the number of antennas along x-axis and y-axis, respectively. There are two main reasons for utilizing the URA structure: (i) it packs a larger number of antennas in a twodimensional (2D) grid under the physical-limited space requirements in practical applications, (ii) it enables three-dimensional (3D) beamforming by employing both azimuth and elevation domains[2],[5]-[7],[12],[16] 3. Since the analog RF beamformer is designed via the AoD information, it is reasonable to assume that each subcarrier experiences a similar AoD support (i.e., mean azimuth/elevation AoD and their spread).…”

“…In[16], it is shown that the AoD parameters (i.e., mean and spread) can be efficiently obtained via an offline deep learning and geospatial data-based estimation technique instead of the traditional online channel sounding.…”