Advancing Ultra-Reliable 6G: Transformer and Semantic Localization Empowered Robust Beamforming in Millimeter-Wave Communications

The reconfigurable intelligent surface (RIS) is promising in fulfilling the requirement of green communication in the fifth- and sixth-generation (5G/6G) mobile communication networks. Considering the limits of the conventional channel state information (CSI) acquisition techniques for the RIS-aided communication system, this paper proposes a location-aware beamforming design for the RIS-aided millimeter-wave (mmWave) communication system without the channel estimation process. First, according to the system geometry, the relationship between the channel model and the user location is established, which is further used to derive the CSI error bound using the location error bound. Then, a worst-case robust beamforming optimization problem is formulated and solved to combat the effect of location error on the beamforming design. Simulation results demonstrate that the proposed approach outperforms the conventional non-robust approach in terms of the worst-case signal-to-noise ratio (SNR) at the receiver, and becomes more advantageous in the presence of higher level of the user location uncertainty.

Millimeter-wave (mmWave) communication offers rich spectrum resources and acts as a key enabling technology for future wireless communication systems. Hybrid (digital and analog) beamforming and relay techniques are important for mmWave implementations considering the characteristics of mmWave signals, as well as the practical limitations of equipment size, power consumption, and hardware cost. In this paper, a robust hybrid beamforming scheme is presented for mmWave multiple-input multiple-output relay networks adopting amplify-and-forward strategy at the relays. Unlike most existing designs that are based on the perfect channel-state information (CSI), CSI imperfectness is considered in the proposed robust beamforming scheme. An accurate approximation of the average received signal-to-noise ratio is derived and used as the design criterion for the developed iterative beamforming optimization at different nodes. An orthogonal matching pursuit-based algorithm is then utilized to design the hybrid beamforming schemes. Simulation results show that the proposed robust beamforming scheme with affordable computational complexities provides substantial performance gains compared with the existing non-robust designs.

The large overhead arising from conventional channel estimations in reconfigurable intelligent surface (RIS) aided millimeter-wave communication systems, may offset the performance gain brought by the RIS. To tackle this issue, we propose a location information assisted beamforming design without the requirement of the channel training process. First, we establish the geometrical relationship between the channel model and the user location, and mathematically derive an approximate channel state information (CSI) error bound based on the user location error region. Then, for combating the negative impact of the location error on the communication performance, we formulate a worst-case robust beamforming optimization problem to optimize the beamformer at the base station (BS) and the phase-shift matrix at the RIS. To solve this non-convex problem, we develop a novel relaxed alternating optimization process (RAOP) by utilizing various optimization tools, such as the Lagrange multiplier, the matrix inversion lemma, the semidefinite relaxation (SDR), as well as the branch-and-bound (BnB). Additionally, we prove sufficient conditions for the SDR to yield rank-one solutions, and modify the BnB to acquire the phase-shift solution under an arbitrary constraint of possible phase-shift values. Finally, we analyse the convergence and complexity of the proposed RAOP, and carry out simulations for performance evaluations. Compared to the conventional non-robust beamforming, our method performs better and shows strong robustness against the location-error-related CSI uncertainty. Compared to the robust beamforming based on the S-procedure and penalty convex-concave procedure (CCP), our method with BnB shows the advantages of being able to converge faster and handle arbitrary phase-shift argument sets.