| dc.description.abstract | Next-generation wireless systems aim to revolutionize communication by achieving data rates up to terabits per second while also supporting diverse applications such as autonomous mobility, industrial automation, and extended reality. With capabilities like millimeter-level positioning and ultra-low latency (sub-millisecond), 6G offers enhanced sensing, enabling technologies such as fully autonomous mobility systems. To meet these performance demands, 6G systems are expected to operate in higher frequency bands (e.g., millimeter-wave and sub-terahertz) with large antenna arrays. However, these advancements face challenges, including the cost and complexity of traditional radio-frequency (RF) chains, harsh propagation environments, and introduces the necessity to enable seamless coexistence of communication and sensing systems.
Coarse quantization and antenna selection are effective methods for reducing RF-chain complexity in wireless systems. Coarse quantization employs low-resolution (often 1-bit) converters to lower cost, complexity, and power consumption, while antenna selection dynamically activates a subset of antennas based on channel conditions. Reconfigurable intelligent surfaces (RIS), composed of tunable meta-material elements, improve wireless propagation by introducing additional paths, enabling reliable communication and sensing even in challenging environments. Integrated sensing and communication (ISAC) systems share hardware and spectral resources, using advanced precoder designs to support the coexistence of communication and sensing, making them a key enabler for 6G.
The adoption of emerging technologies such as coarse quantization, antenna selection, RIS, and ISAC significantly alters traditional wireless transceiver design. Conventional precoding and channel estimation algorithms, developed under the assumption of full-resolution quantization, suffer severe performance degradation in presence of quantization and antenna selection. Additionally, RIS and ISAC introduce entirely new design complexities, rendering existing signal processing methods inadequate. In this thesis, we address these challenges by developing appropriate algorithms and signal processing techniques.
In the first part of the thesis, we focus on communication systems. We begin by considering channel estimation in MIMO systems equipped with 1-bit spatial sigma-delta quantizers—a technique that leverages spatial oversampling and feedback. We develop a novel quantization noise model and propose a parametric channel estimation algorithm to estimate the millimeter-wave communication channel by estimating the associated angles and path gains. Additionally, we introduce a method to select the quantization voltage levels and sigma-delta feedback coefficients. Our proposed approach enables spatial sigma-delta quantized systems to achieve performance comparable to unquantized systems while significantly outperforming conventional 1-bit quantized systems.
Next, we consider an RIS-assisted communication system. To tackle challenges in harsh propagation environments, we propose a novel RIS architecture comprising both active and passive elements for next-generation communication systems. We develop a low-complexity solver to optimize RIS coefficients and to determine the placement of active elements in a channel-aware manner. This approach achieves improved signal-to-noise ratios compared to existing schemes that rely solely on fully passive or fully active RIS architectures.
In the second part of the thesis, we focus on ISAC systems. We first examine a simple RIS-assisted ISAC scenario involving a single user and a single target. To simultaneously localize the target and serve the user, we propose an adaptive RIS partitioning strategy, allocating distinct elements for sensing and communication. We demonstrate that a shared RIS can significantly enhance both communication and sensing performance. We then extend this to a more complex setting with multiple users and multiple targets. We develop algorithms to jointly design transmit precoders and RIS coefficients that maximize sensing performance while maintaining the quality of service for communication users across various scenarios. By using an appropriately designed RIS, we demonstrate that ISAC systems can achieve the desired communication and sensing performance even when the direct link from the base station to the users and the targets are completely blocked.
Next, we address the issue of RF-chain complexity in ISAC systems. We propose a novel antenna selection method that benefits from diversity for communication while still guaranteeing the radar system the ability to estimate the directions of a certain number of targets. Specifically, we present an algorithm to perform antenna selection and transmit precoding for ISAC systems that guarantee both communication and sensing requirements.
Until now, we have explored traditional optimization-based methods for ISAC precoding. Optimization-based precoding methods usually result in high computational complexity and suffer from performance deterioration in the presence of channel estimation errors. To address these shortcomings, we present a data-driven method for ISAC precoding. Specifically, we present an unsupervised learning model to learn the precoders directly from the received pilots and echoes, thereby avoiding the need for explicit channel estimation. We present a neural network architecture that works for varying users and targets. We further propose a novel loss function based on first-order optimality conditions to train the neural network in such a way that the output of the neural network not only results in improved sensing performance but are also likely to satisfy non-convex communication SINR constraints. Compared to traditional optimization-based methods, the proposed learning-based approach delivers superior performance with significantly reduced computational complexity.
Thus far, we have focused on standalone ISAC systems with a single dual-function radar-communication base station (DFBS). However, in practical scenarios involving large geographic areas, multiple DFBSs are required. The coexistence of standalone DFBSs introduces inter-cell interference (ICI), which must be effectively managed. Distributed ISAC (DISAC) is a promising approach to address this challenge, enabling multiple interconnected DFBSs to serve users and targets collaboratively via a shared backhaul link. However, Existing DISAC designs often require excessive transmit power or high backhaul rates. We propose a novel cooperative transmission strategy for DISAC systems that balances power and backhaul rate tradeoffs to address these limitations. Our approach reduces backhaul requirements by selectively designing the subset of base stations (BS) serving each user (UE). We develop transmit precoders to minimize a weighted sum of total transmit power and backhaul rate while ensuring the required signal-to-interference-plus-noise ratios (SINRs) for both communication and radar sensing. We present a convex-relaxation-based method to obtain the transmit precoders. For millimeter-wave DISAC systems with large antenna arrays, we further introduce a lower-complexity alternative. By leveraging the line-of-sight nature of mmWave channels, we simplify the precoding problem to spatial window design and power allocation problems, for which we provide a computationally efficient solution. | en_US |
