About LiNKS – Laboratory for Intelligent NetworKed Systems
Our lab conducts forefront research in wireless communication theory, developing advanced communication and network algorithms for next-generation wireless systems, including beyond 5G and 6G networks. We tackle complex challenges including:
- Efficient resource allocation in multiuser and multiagent systems
- Optimized scheduling, associations, and handovers in dynamic networks
- Advanced beamforming techniques for improved transmission and reception
- Cross-layer optimization integrating constraints and interactions across hardware, network, and other critical layers
To solve these challenges, we employ cutting-edge optimization and machine learning techniques, including:
- Custom algorithm design for non-convex, non-linear, and mixed-integer problems
- Deep reinforcement learning strategies for real-time, multi-agent systems
- Graph neural networks for capturing network interactions and improving generalizations
We welcome collaboration with experts in complementary fields, including communications, networking, experimental testbeds, and hardware design. Additionally, we are open to exploring opportunities to apply our optimization and machine learning skills to innovative areas beyond wireless communications.
Recent News
- Sept 2025: Welcome to our new PhD student Tien Vu joining us from the Ho Chi Minh City University of Technology (HCMUT).
- July 2025: Welcome Professor Lim for his summer visit to our lab. We are very happy to see you again, Byungju, and looking forward to more collaboration.
- June 2024: Prof. Vu was invited to participate in the MathWorks Research Summit 2024 on a panel about Reinforcement Learning. Here is the set of opening slides.
- May 2024: Congratulations to Alireza for winning the 2023 Joseph P. Noonan Outstanding Doctoral Research Award! Well done, Dr. Alizadeh!
- Sept. 2023: Welcome to our new PhD student Changgyu Lee who joined us from Kyung Hee University, South Korea.
- Feb. 2023: Byungju has accepted the position as an Assistant Professor at Pukyong National University, Busan, Korea. He will start in March. Congratulations Professor Lim!
- Jan. 2023: Qing Lyu arrived at our lab as a PhD student. Welcome Qing!
- Dec. 2022: Seok-Hyun completed his visit with us and will return to finish his PhD degree at Korea University.
- Nov 2022: Prof. Vu joined the IEEE Transactions on Communications Editorial Board as an Editor.
Recent Publications
- Our paper is accepted to NeurIPS’25, Workshop on AI and ML for Next-Generation Wireless Communications and Networking. In this paper entitled “Towards Achieving Integer and Load-balancing User Association in Wireless Networks with a Reparameterized Attention-based GNN,” we demonstrated novel uses of Gumbel-Softmax reparameterization to meet integer constraints, and Sinkhorn normalization method together with regularization to meet non-linear load balancing constraints. These constraints have been challenging for ML models to satisfy.
- We designed new GNN structures for joint beamforming and user association in a wireless network, with efficient edge representations for beamforming and employing a novel reparameterization method to achieve near integer association outputs. These results, to appear on the IEEE Transactions on Vehicular Technology and be presented at the IEEE Globecom conference, are steps towards using ML in practical engineering systems with constraints.
- Our new collaborative work explores an efficient way to acquire channel state information (CSI) in a multi-RIS aided wireless network that significantly reduces the estimation signaling overheads. By estimating the composite (instead of individual) reflected channels from each RIS and integrating this CSI in between the layers of a GNN model, the learning can adapt to the new CSI seamlessly during the channel estimation process. Our results are reported in a paper recently accepted to the IEEE Transactions on Wireless Communications.
- Our new paper demonstrates the efficiency of using Graph Neural Networks (GNN) for real-time hybrid beamforming in wideband multicarrier systems, minimizing computation overhead. The GNN-based design also effectively mitigates beam squinting effects. These findings will be presented at the 2024 IEEE International Symposium on Phased Array Systems & Technology.
- We proposed multi-agent Q-learning algorithms for real-time user association and handover in dense 5G/6G networks. Centralized and distributed multi-agent policies improve load balancing, reducing handover rates and increasing network throughput. The work is detailed in a paper published in the IEEE Transactions on Wireless Communications.
- Our analysis of 6G coexistence between terrestrial networks and passive satellites provides bounds for node density and transmitted power to achieve near-zero outage, together with conditions for out-of-band interference to meet spectral masks. Results are published in IEEE Transactions on Wireless Communications.