A
Near-field communications & sensing
Low-overhead beam acquisition, refinement, and tracking for large apertures under spherical-wave propagation.
- DFT and Fresnel codebooks
- Amplitude-only measurements
- Beam-pattern analysis
Wireless communications · Reinforcement learning · Hardware prototyping
Zijun
Wang.
王梓骏
Ph.D. student in the Department of Electrical and Computer Engineering at the University at Buffalo
I develop model-driven and learning-assisted signal processing methods for near-field communications and sensing, with applications to beam training and tracking, sparse array design, and mmWave prototyping.
I received my B.Eng. in Telecommunications Engineering from Nanjing University in 2024. I am now pursuing the Ph.D. degree at UB, advised by Prof. Rui Zhang.
- Near-field MIMO
- Sparse inference
- 28 GHz testbeds
- Deep reinforcement learning
01 / News
Recent updates
Selected publication and open-source milestones.
-
Efficient Near Field Beam Tracking via Thompson Sampling has been accepted by the 2026 Asilomar Conference on Signals, Systems, and Computers.
-
Adaptive Payload-Aided Near-Field Beam Tracking via Thompson Sampling is now available on arXiv and indexed by Google Scholar.
-
New articles appeared in IEEE Transactions on Wireless Communications and IEEE Wireless Communications Letters.
-
Two near-field beam-training papers were published at IEEE GLOBECOM 2025, with complete research code released on GitHub.
02 / Research
Research areas
Physics-aware algorithms that connect propagation models, inference, control, and real RF hardware.
B
Array & beamspace signal processing
Structured recovery methods that exploit spatial sparsity and codebook geometry for efficient channel acquisition.
- Multi-beam combination
- Total variation denoising
- Off-grid refinement
C
Learning & cross-layer optimization
Bandit and reinforcement-learning methods for beam alignment, tracking, scheduling, and delay-sensitive traffic.
- Thompson sampling & UCB
- Linear & GP bandits
- PPO & cross-layer scheduling
D
Wireless systems & testbeds
End-to-end 28 GHz prototyping spanning waveform generation, synchronized capture, beam control, and evaluation.
- USRP X410 and UHD
- TMYTEK 8×8 phased arrays
- MATLAB / Python automation
03 / Publications
Selected publications
Journal articles
Conference & workshop papers
2026 Asilomar Conference on Signals, Systems, and Computers
Efficient Near Field Beam Tracking via Thompson Sampling
NeurIPS 2025 AI4NextG Workshop · OpenReview
Cross-Layer Design for Near-Field mmWave Beam Management and Scheduling under Delay-Sensitive Traffic
Preprints listed on Google Scholar
arXiv:2603.07477
GP Bandit-Assisted Two-Stage Sparse Phase Retrieval for Amplitude-Only Near-Field Beam Training
04 / Projects
Open-source projects & testbeds
01
View repository
02
View repository
03
View repository
04
View repository
05
Amplitude-only beam training & localization
Precise Near-Field Beam Training with DFT Codebook Based on Amplitude-Only Measurement
Low-Complexity Near-Field Beam Training with DFT Codebook Based on Beam Pattern Analysis
Sparsity-aware multi-beam training
Sparsity-Aware Near-Field Beam Training via Multi-Beam Combination
Cross-layer beam management & scheduling
Cross-Layer Design for Near-Field mmWave Beam Management and Scheduling under Delay-Sensitive Traffic
Compressive LASSO-TV beam training
Compressive Beam-Pattern-Aware Near-Field Beam Training via Total Variation Denoising
28 GHz communication & sensing platform
End-to-end mmWave prototyping
Experience integrating a USRP X410, UHD 4.8, TMYTEK 8×8 phased arrays, YTTEK frequency converters, horn antennas, and laboratory instrumentation.
View example repository