The metaverse is expected to provide immersive entertainment, education, and business applications. However, virtual reality (VR) transmission over wireless networks is data- and computation-intensive, making it critical to introduce novel solutions that meet stringent quality-of-service requirements. With recent advances in edge intelligence and deep learning, we have developed a novel multi-view synthesizing framework that can efficiently provide computation, storage, and communication resources for wireless content delivery in the metaverse. We propose a three-dimensional (3D)-aware generative model that uses collections of single-view images. These single-view images are transmitted to a group of users with overlapping fields of view, which avoids massive content transmission compared to transmitting tiles or whole 3D models. We then present a federated learning approach to guarantee an efficient learning process. The training performance can be improved by characterizing the vertical and horizontal data samples with a large latent feature space, while low-latency communication can be achieved with a reduced number of transmitted parameters during federated learning. We also propose a federated transfer learning framework to enable fast domain adaptation to different target domains. Simulation results have demonstrated the effectiveness of our proposed federated multi-view synthesizing framework for VR content delivery.

Geosynchronous spaceborne-airborne bistatic synthetic aperture radar (GEO-BiSAR) has the advantages of wide beam coverage, long exposure time, and superior system flexibility. However, it is a challenge for GEO-BiSAR to efficiently acquire SAR images both with high resolution and wide swath due to the severe range and azimuth spatial variances. To deal with this issue, a modified Omega-K method is proposed in this paper. In the proposed algorithm, an equivalent range model associated with GEO-BiSAR configuration is built, where the equivalent parameters of the airborne radar receiver absorb the phase parameters corresponding to the GEO transmitter. Then, the Stolt interpolation is performed to realize the range-azimuth decoupling, achieving the linearization between range and frequency variables. Finally, a well-focused SAR image could be obtained after residual spatial variance error compensation. Simulations are presented to demonstrate the effectiveness of the proposed method.

Digital immersion via virtual reality (VR) has promising applications in entertainment, education, and business. However, VR transmissions over wireless are data-intensive and computation-intensive. It is critical to investigate novel wireless network solutions that meet stringent quality-of-service requirements in VR. In this paper, we propose a novel VR transmission scheme to transmit single-view images only. Particularly, single-view images are broadcasted to users with the overlapped field of view, and the corresponding multi-view consistent content is generated by a neural network to avoid massive content transmission. We design a federated learning framework to guarantee an efficient learning process by characterizing vertical and horizontal data samples. Meanwhile, exchanging parts of models during the federated learning process can achieve low-latency communications. Simulation results validate the effectiveness of the proposed VR transmission scheme.

This paper considers a two-user downlink transmission in intelligent reflecting surface (IRS) aided network over fading channels. Particularly, non-orthogonal multiple access (NOMA) and two orthogonal multiple access (OMA) schemes, namely, time division multiple access (TDMA) and frequency division multiple access (FDMA), are studied. The objective is to maximize the system average sum rate for the delay-tolerant transmission. We propose two adjustment schemes, namely, dynamic phase adjustment and one-time phase adjustment. The power budget, minimum average data rate, and discrete unit modulus reflection coefficient are considered as constrains. To solve the problem, two phase shifters adjustment algorithms with low complexity are proposed to obtain near optimal solutions. With given phase shifters and satisfaction of time-sharing condition, the optimal resource allocations are obtained using the Lagrangian dual decomposition. The numerical results reveal that: i) the average sum rate of proposed NOMA network aided by IRS outperforms the conventional NOMA network over fading channels; ii) with continuous IRS adjustment in the fading block, the proposed TDMA scheme performs better than the FDMA scheme; iii) increasing the minimum average user rate requirement has less impact on the proposed IRS-NOMA system than on the IRS-OMA system.

This paper considers a two-user downlink intelligent reflecting surface (IRS) assisted network over fading channels. Particularly, non-orthogonal multiple access (NOMA) and two orthogonal multiple access (OMA) schemes, namely, time division multiple access (TDMA) and frequency division multiple access (FDMA), are studied. Our goal is maximizing the system average sum rate for delay-tolerant transmission. The power budget, minimum average user rate and discrete unit modulus reflection coefficient constraints are considered as constraints. To solve the problem, phase shifters adjustment algorithms with low complexity are proposed to obtain high quality solutions. With given phase shifters, the optimal power allocation is obtained using the Lagrangian dual decomposition. Numerical results demonstrate the performance gain by introducing IRS into the system as well as the utility of the proposed algorithms.