Session: Optimization Methods for Next-Generation Wireless Communication Networks (I)
Chair: Ya-Feng Liu
Cluster: Optimization Applications (Communication, Energy, Health, ML, ...)
Talk 1: Exploiting Statistical Hardness for Private Wireless Localization
Speaker: Urbashi Mitra
Abstract: Securing signals from unintended eavesdroppers has become an increasingly important problem in next generation (5G or 6G) wireless communications with the emergence of the Internet-of-Things and Machine-to-Machine communications. Herein, we examine learning problems in signal processing that are inherently hard without key side information. In particular, we exploit necessary resolution limits for classical compressed sensing/sparse approximation problems. To limit an eavesdropper's capabilities, we create an environment for the eavesdropper wherein the appropriate structured statistics algorithm would provably fail. The intended receiver overcomes this ill-posed problem by leveraging a very modest amount of secret side information shared between the intended transmitter and receiver. Two instantiations of private localization are considered, both independent of channel state information and both based on the design of a novel precoder at the transmitter of the device to be localized. In the first case, spurious, virtual multipath is introduced so that the eavesdropper perceives a much more rich channel than the intended user. In the second case, the channel perceived by the eavesdropper has multipath that appears to have been moved from the original, resulting in the eavesdropper learning a spoofed location. Parameter design is enabled through the development of bounds on the estimation error for the eavesdropper. We pose optimization problems whose solutions yield parameter designs for maximal security. Proper parameter design can result in a significant increase (orders of magnitude) in the localization error for the eavesdropper. Theoretical guarantees are provided for all problems considered. All proposed algorithms are validated via numerical results, and it is seen that the eavesdropper’s capabilities are severely degraded.
Talk 2: Reconfigurable Intelligent Surface-aided Wideband Communications
Speaker: Chandra Murthy
Abstract: In this talk, we discuss the beam-split effect in wideband communications aided by a large passive reconfigurable intelligent surface (RIS) equipped with configurable phase shifters. The beam-split is caused by the spatial delay spread across the large aperture of the RIS coupled with its phased array architecture, and results in different frequency components of the wideband signal getting beamformed to different directions. As a result, only a few subcarriers of an orthogonal frequency division multiplexing (OFDM)-based transmission get correctly reflected by the RIS towards a desired user. In turn, this severely degrades the achievable data rate. We present two approaches to alleviate the beam-split effect. The first is a distributed RIS approach where the size of the individual RISs is chosen to ensure that the effect of beam-split at each RIS remains within a tolerance level. The second approach exploits beam-split instead of controlling it, by opportunistically allotting non-overlapping sub-bands of the wide bandwidth to different users for a given RIS phase configuration, thereby leveraging multi-user diversity. The former case provides near-optimal RIS benefits for a given user but requires careful RIS placement to control the temporal delay spread introduced by the multiple RISs. The latter case provides near-optimal benefits in terms of the network throughput but requires many users in the system. We contrast the two approaches via numerical simulations. This is joint work with Yashvanth L.
Talk 3: Fractional Programming for Discrete Optimization in Signal Processing and Machine Learning
Speaker: Kaiming Shen
Abstract: Fractional programming (FP) refers to optimization problems involving functions of ratios. FP plays an important role in signal processing and machine learning, because many problems in these application areas are fractionally structured. This talk focuses on a state-of-the-art FP method called the quadratic transform and illustrates the use of quadratic transform for discrete optimization problems through two application examples. The first example is the joint optimization of beamforming and user scheduling for uplink wireless cellular networks, which is a mixed discrete and continuous nonlinear optimization problem. In this domain, the weighted minimum mean square error (WMMSE) algorithm has been extensively used. We connect WMMSE to FP, and further improve upon WMMSE by using the quadratic transform. The second example is the 0-1 normalized-cut (NCut) problem for data clustering and image segmentation. Unlike the previous relaxation-based methods, the proposed FP method recasts the NCut problem into a sequence of weighted bipartite matching problems, which can be solved efficiently without relaxing the discrete variables.
