Abstract
The 6G wireless communication network has shown its tremendous advantages in digital transformation of societies. 6G enables reliable, pervasive, and near instant wireless connectivity by integrating aerial, maritime, and terrestrial communications. The development of cutting-edge technologies like machine learning, blockchain, millimeter wave communication, non-orthogonal multiple access, tera-Hertz communication, quantum communication/quantum machine learning, fog/edge computing, and tactile Internet encourages the demand of beyond 5G and 6G communication. An effective wireless communication system must be capable of satisfying the user requirements such as increase in capacity, efficiency flexibility, coverage, and quality of experience. As the number of users and coverage area increases, the complexity of designing a 6G network is also getting increased. In recent years several research works are applying ML to wireless communication. This chapter discusses about the application of ML in the area of massive MIMO, NOMA,OWC, polar codes, and security of wireless communication.
Top2 Massive Mimo
Rapid growth in the number of cellular users imposes the need for new emerging technologies to satisfy the user demands. The massive multiple input multiple output (mMIMO) system is a key technology for boosting 5G networks' user capacity. In millimeter wave (mmWave) systems with greater complexity and physical testing, mMIMO performs well (Alkhateeb & Beltagy, 2018; Lamare, 2013; Liu et al., 2016).
Machine learning methods are being applied to both mmwave and mMIMO in an effort to simplify real-time implementation (Liu, 2014; Zappone et al., 2018). ML delivers higher performance with less complexity when compared to the existing techniques, such as game theory, stochastic geometry, and combinatorial optimization. Machine learning is important in MIMO load balancing, beamforming, spectrum optimization, and channel prediction due to its dynamic nature (Jiang et al., 2017; Simeone, 2018).
To integrate ML in a larger mMIMO system, the developers of (Booth, 2019) used software defined radio. They implemented the hardware framework and examined the speed of huge MIMO simulations using the LimeSDR development kit and the Lime Suite signal processing environment. The findings indicate that the bottlenecks between the base station (BS) and user are reducing the throughput and dependability of mmWave (Alkhateeb & Beltagy, 2018). The IRES project uses stationary transmitters and receivers with a moving obstruction to get around this. In order to follow the previous beamforms, a gated recurrent unit (GRU) is implemented, and the likelihood that the following beamform will be blocked is determined. In order to obtain the desired results, the system is first simulated using the DeepMIMO dataset (Alkhateeb, 2019). The data is then utilised to test and train the recurrent neural network (RNN).