Efficient characterization and optimization of smart wireless environments

Ross, Charles

Permalink

https://hdl.handle.net/2142/129720 Copy

Description

Title

Efficient characterization and optimization of smart wireless environments

Author(s)

Ross, Charles

Issue Date

2025-04-29

Director of Research (if dissertation) or Advisor (if thesis)

Peng, Zhen

Doctoral Committee Chair(s)

Peng, Zhen

Committee Member(s)

• Bernhard, Jennifer
• Schutt-Aine, Jose
• Moon, Thomas

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Reconfigurable Intelligent Surface
• MIMO
• Multi-User MIMO
• Bayesian Optimization
• Channel Model
• Ising Hamiltonian
• Quantum Annealing
• High-Dimensional Optimization
• Discrete Optimization

Abstract

Reconfigurable intelligent surfaces (RIS) and Multi-user MIMO (MU-MIMO) have emerged as promising technologies for enhancing the system through- put, channel capacity, and energy efficiency of wireless networks, transforming static propagation environments into dynamic spaces that may adapt in real time. These technologies introduce a large number of tunable parameters, creating a high-dimensional design space that presents significant computational challenges. Fully realizing their potential requires 1) learning the non-linear dependency of these parameters and 2) efficiently optimizing the learned function. In this thesis, we will describe different frameworks that address these challenges. To begin, the equivalence between RIS optimization and Ising Hamiltonians is explored. With this formulation of the problem, RIS may be optimized using quantum and quantum-inspired algorithms. This representation captures a wide range of RIS use cases, including multiple receivers, multipath environments, and distributed RIS. We next consider the challenge of characterizing the functional dependence. Based on the Ising representation, a novel algorithm, tensor contraction with regression (TCR), is used to learn the end-to-end channel. TCR enables rapid convergence, with minimal pilot overhead, to the ideal RIS phase configuration. Next, we explore a Bayesian Optimization (BO) framework equipped with physics-informed dictionary embedding. This approach eliminates the need for explicit channel estimation and enables sample-efficient optimization of RIS configurations. We extend this framework with a fully Bayesian surrogate model using structured priors, facilitating joint optimization of RIS and precoding vectors in MU-MIMO systems. Our method effectively addresses the mixed discrete-continuous nature of the design space, demonstrating superior performance across diverse wireless scenarios while significantly reducing computational complexity compared to conventional approaches.

Graduation Semester

2025-05

Type of Resource

Thesis

Handle URL

https://hdl.handle.net/2142/129720

Copyright and License Information

Copyright 2025 Charles Ross

Owning Collections