Abstract
Future wireless systems are expected to evolve towards an intelligent and softwarebased functionality paradigm, where wireless networks are not only allowing people, mobile devices, and objects to communicate with each other, but also capable of sensing, controlling, and optimizing the environment to realize the vision of the vision of low powered, high throughput and low latency communications. To achieve this goal, we proposed a novel wireless communication framework named as Reconfigurable Intelligent Surface (RIS), a transformative and emerging wireless technique based on the programmable meta-surface. In this thesis, we first provide a transformative framework for future wireless communications, by introducing its basic hardware architectures and main characteristics, highlighting the key design considerations as well as the new opportunities to be exploited. As we all know, every new technology is accompanied by new challenges. The thesis considers three main components of RIS-based wireless communication networks: 1) Channel Estimation for RIS-based Millimeter wave (mmWave) MIMO Systems; 2) Resource Allocation and Energy Efficiency for RIS-based Wireless Systems 3) Non-convex Configuration for RIS-based Systems. The correspondingly research focuses and contributions were summarized as follows. RIS-based mmWave communication can provide higher bandwidth and communication rates than any currently commercialized network. Because it operates at extremely high frequencies (30GHz-300GHz), this allows base stations and clients to integrate large-scale antenna arrays to overcome its natural communication weaknesses- severe channel degradation. This inevitably leads to the communication channel having the sparse characteristics. The traditional channel estimation methods are not suitable for mmWave communication. To solve this problem, the first focus is to propose low complexity iterative algorithms to for RIS-based mmWave MIMO systems. Regardless of the specific implementation, what makes the RIS technology attractive from an energy consumption standpoint, is the possibility of amplifying and forwarding the incoming signal without employing any power amplifier. As far as resource allocation and Energy Efficiency (EE) are concerned, it is not clear if a RIS-based system is more convenient than traditional relay-based systems. The second components aim at answering this question, showing that by properly designing the phase shifts applied by the RIS leads to higher EE than that of a communication system based on traditional relays. For the increasingly complex distributed networks in the future, as well as the need for intelligent sense, the wireless communication channel is generally extreme complex and has non-linear characteristics. In addition, it is a thorny issue for the phase configuration in distributed intelligent communication systems. Combining above-mentioned issues together, the third important focus is to propose a generalized deep neural network to solve the problem of non-convex channel estimation and optimal phase configuration in the distributed RIS-based complex networks.