PhD defence of Alexander Fernandes – Reconfigurable intelligent surface-assisted wireless communication systems: signal processing techniques for channel estimation and phase shift compression

Abstract

"The reconfigurable intelligent surface (RIS) is composed of passive reflective elements designed to reconfigure the wireless propagation environment for the next generation of wireless communication systems. In RIS-assisted wireless communications, the main goals include channel estimation to acquire the channel state information (CSI), and joint beamforming and phase shift reconfiguration to improve the achievable data rates. Due to the design challenges of making passive reflective elements, one challenge is to deal with hardware impairments (HIs) on the desired phase shifts during signal propagation. Another challenge is that the number of parameters to estimate increases with the number of passive elements, thereby introducing a larger channel estimation overhead trade-off between pilot training duration and channel estimation accuracy. After channel estimation, knowledge of the CSI will be acquired at the access point (AP), this poses another challenge as the RIS controller that updates the phase shifts relies on information transfer from the AP to the RIS, introducing a communication overhead scaling linearly with number of elements. The objective of this dissertation is to address these challenges in RIS-assisted wireless communications.

The first study is on channel estimation in a full-duplex (FD) wireless communication system assisted by a RIS with HI occurring at the transceivers and RIS elements. We propose an unbiased channel estimator that requires knowledge of only the first and second order statistics of the HI, for which we derive closed form expressions. The proposed estimator reduces to the maximum likelihood estimator in the case of ideal hardware. We also describe simultaneous and non-simultaneous orthogonal pilot schemes that minimize the mean square error of the maximum likelihood estimator in the case of ideal hardware.

The next studies introduce a novel tensor signal model for channel estimation of a RIS-assisted communication model for half-duplex (HD), which is then extended to FD. For the HD and FD models, we use tensor signal modelling techniques to estimate all CSI involving the self-interference, direct-path, and the RIS assisted channel links. We model the received signal as a tensor composed of two CANDECOMP/PARAFAC (CP) decomposition terms for the non-RIS and the RIS assisted links, extend the alternating least squares algorithm to jointly estimate all channels, then derive the corresponding Cramér-Rao Bound (CRB). The proposed method provides a more accurate estimate by efficiently using all pilots transmitted throughout the full training duration without turning the RIS “OFF” when comparing the same number of total pilots transmitted. For a sufficient number of transmitted pilots, the proposed method’s accuracy comes close to the CRB for the RIS channels and attains the CRB for the direct-path and self-interference channels.

The final study is on the design of a deep learning (DL) architecture to implement joint phase shift compression and beamforming using knowledge of the acquired CSI. We propose a model-based DL architecture to reduce the number of bits required for transmitting phase shift information from the AP to the RIS controller. The AP computes the phase shifts and compresses them into a binary control message that is sent to the RIS controller for element configuration. To help reduce beamformer mismatches caused by phase shift compression errors, the beamformer is updated with the actual (decompressed) RIS phase shifts. By unrolling the iterative weighted minimum mean square error (WMMSE) algorithm within the wireless communication-informed DL architecture, joint phase shift compression and beamforming can be trained end-to-end. The proposed model-based DL architecture demonstrates that incorporating compression-aware beamforming significantly improves sum-rate performance, even when the number of control bits is lower than the number of RIS elements.