Executive Summary : | Massive MIMO has been state-of-the-art 5G technology for building the basestations. It consists of many active antenna elements that can direct the signals up and down from the transmitter in the base stations. After almost 10 years of research, recently, SPRINT deployed this technology at thousand of sites in USA [1]. Their configuration contains 64 antennas in the 2.5 GHz band improving the data rate 4 to 20 times. But still it is far away from what people have been talking about in the academic research for last many years where we wanted to have millimeter long arrays. This is due to the problem in its practical deployment of base antennas. In order to build the large arrays for future 6G communications, without the expense of multiple antennas or incurring large energy consumption, metasurfaces also known as reconfigurable intelligent surfaces (RIS)/software-controlled metasurfaces [2–4] have recently emerged as a potential candidate to enhance the coverage and capacity. The metasurafces contain small scattering meta-atoms of sub-λ size (λ-wavelength) which scatters the incoming wave with controllable delay leading to controllable phase shift. The signal coming from a RF transmitter bounces of the RIS with a controllable bean direction as well as controlled shape and reaches the receiver. This in return, helps in controlling the directivity of the scattered signal, signal absorption and polarization. It leads to improved indoor coverage, protection against eavesdropping and mitigation of the shadow fading. These surfaces can be deployed in the rural (village) areas where power availability is a constraint as required by the traditional AF relay networks. In this project, we target to reinforcement learning algorithms for RIS technology which will be proved a motivation for the deployment of RIS technology especially in the rural areas. In addition, we target to develop efficient algorithms for the channel estimation with the minimum pilot overhead. We will propose the theoretic framework for the developed algorithms in addition to their validation in the real radio environment. |