Project Title: Quantum and Quantum-Inspired Computing for Large-Scale NOMA-MIMO Wireless Networks
WINLAB Summer Internship 2025
Group Members: Alexander MarkleyGR, Jeffrey TangUG, Shehzad SultanUG, Brian ChaconUG, Oskar IwaniukUG
Advisors: Minsung Kim, Byungjun Kim
Project Objective
The project will explore non-traditional computing methods for “Non-Orthogonal Medium Access-based Multiple-Input Multiple-Output” (NOMA-MIMO) wireless systems. Both MIMO and NOMA are considered among the most promising techniques to increase wireless capacity by scaling up the number of serviced devices at a time. However, to do so, they require much more computationally demanding processing at the receiver. A proposed solution is to reduce MIMO Maximum Likelihood Detection (MLD) to Quadratic Unconstrained Binary Optimization(QUBO), which resembles a Hamiltonian. We, then, convert QUBO into the Ising form under the Ising model, and use an Ising solver for the best Ising configurations. Finally, the best candidate will be mapped to MIMO Detected Bits.
Over the summer, our team split into three groups to handle …
Weekly Progress
WEEK ONE
Week 1 NOMA-MIMO Presentation
Week 1 LDPC
Week 1 Wifi Based Detection Presentation
Progress:
- Familiarize ourselves with OrbitLab, ParaMax: quantum-inspired algorithm using simulated annealing and parallel tempering for MIMO ML detection, and other state-of-the-art quantum approaches for NOMA-MIMO
- Researched Wifi-Fingerprinting and Human Adversarial Attack, and continued Integer optimization (Brian add here)
WEEK TWO
Week 2 NOMA-MIMO Presentation
Week 2 LDPC Presentation
Week 2 Wifi Based Detection Presentation
Progress:
- Completed tutorials for GNURadio, USRP2, X310; Reviewed wireless packet detection and synchronization; Reviewed ParaMax architecture
- Followed tutorial to rebuild a Wifi Fingerprinter for bit similar devices; used a Convolutional Neural Network for this
- Edited the quadratic unconstrained binary optimization (QUBO) into a discrete solver using integer variables
WEEK THREE
Week 3 NOMA-MIMO Presentation
Week 3 LDPC
Week 3 Wifi Based Detection Presentation
Progress:
- Obtained access to the Grid; Successful SISO transmission from nodes to MIMO rack
- Fine-tuned and debugged the integer solver to correctly calculate the problem's energy
- Familiarized with Orbit and ILabs to run CNN training
WEEK FOUR
Week 4 NOMA-MIMO Presentation
Week 4 LDPC
Week 4 Wifi Based Detection Presentation
Progress:
- Began looking at Packet Carrier Frequency Offset(CFO) correction and Channel Estimation in MATLAB; Investigating UHD integration and construction of GNURadion Out-Of-Tree (OOT) custom C++/Python blocks
WEEK FIVE
Week 5 NOMA-MIMO Presentation
Week 5 LDPC
Week 5 Wifi Based Detection Presentation
Progress:
- Ran into difficulties with Packet CFO correction, achieved a 37.5-50% channel estimation. Also, due to the ceiling nodes having no central node for synchronization, we were faced with distributed systems communication issues. This led to a shift from MIMO rack to MIMO rack transmission and receiving. We began investigating FlexCore and Multisphere detection as alternative classical parallelized detectors.
WEEK SIX
Week 6 NOMA-MIMO Presentation
Week 6 LDPC
Week 6 Wifi Based Detection Presentation
Progress:
- Began software-based implementation of FlexCore and continued GNURadio setup verification for MIMO rack to MIMO rack transmission
WEEK SEVEN
Week 7 NOMA-MIMO Presentation
Week 7 LDPC
Week 7 Wifi Based Detection Presentation
Progress:
- Adjusted settings for software testing of FlexCore and other conventional MIMO detectors (ZF, MMSE, FCSD, ML)
- Verifying time synchronization on the transmitter nodes
- LDPC: Used Bayesian optimization to initially tune, saw results, and then included more parameters that may be having an impact. Did research into how the model works and what parameters would have the most impact.
WEEK EIGHT
Week 8 NOMA-MIMO Presentation
Week 8 LDPC Presentation
Week 8 RF Digital Fingering Presentation
Progress:
- Ran into performance issues for FlexCore (not reaching near-ML performance), so we started unit testing on the FlexCore methods to ensure correctness and consistency with the details of the FlexCore paper
- LDPC: Plotted first SNR vs. BER plot, going to change parameters and try plotting again. Integrated larger examples into the model, using python ldpc generator script
WEEK NINE
Week 9 NOMA-MIMO Presentation
Week 9 LDPC Presentation
Week 9 RF Digital Fingering Presentation
Progress:
WEEK TEN
NOMA-MIMO Final Presentation
LDPC Final Presentation
RF Digital Fingering Final Presentation
Summary:
Acknowledgements
We want to thank Minsung Kim and Byungjun Kim for their guidance throughout the summer. We also thank Jennifer Shane, Ivan Seskar, and the WINLAB faculty and staff for their support.
