Twelve papers accepted across ICASSP and ICC highlight momentum in wireless research

Twelve research papers from the Centre for Wireless Communications, University of Oulu, have been accepted to two major IEEE conferences in 2026, IEEE ICASSP and IEEE ICC. Taken together, the papers offer a clear snapshot of where learning-driven wireless research is heading, and how integrated sensing, communications, and distributed architectures are beginning to converge.

ICASSP and ICC address different layers of the wireless stack. ICASSP focuses on signal processing foundations, while ICC places emphasis on communication systems and networks. Both conferences are highly selective, and multiple acceptances from the same research environment in a single year are uncommon. Seeing related work appear across both venues in the same cycle reflects a research direction that spans theory, algorithms, and system-level design.

“What matters is not the number of papers, but the kind of questions they are addressing,” says Nhan Nguyen, Assistant Professor and Academy Research Fellow at the University of Oulu and 6G Flagship. “Seeing work from the same research environment appear across both ICASSP and ICC in the same year suggests that ideas developed at the signal level are being tested against system-level constraints.”

When sensing and communication start to converge

Several of the accepted papers focus on integrated sensing and communications (ISAC). Instead of treating wireless transmission and environmental sensing as separate functions, these systems use the same signals to do both. The result is a network that adjusts its behaviour based on what it can observe, rather than one that operates on fixed assumptions.

A persistent difficulty in integrated sensing and communications is change. Beam directions shift, channels fluctuate, and sensing targets move, often faster than static optimisation methods can respond. The ICASSP papers explore learning-based approaches that allow systems to adjust continuously to these conditions. Methods such as reinforcement learning, attention learning, and knowledge distillation help the network track beams, anticipate channel behaviour, and coordinate sensing and communication more effectively over time at minimal computational and power costs.

Rather than treating sensing as an add-on, the work positions it as a driver of communication decisions. That distinction matters in practice. Real networks operate with incomplete feedback, interference, and uncertainty, conditions under which tightly coupled sensing and communication can improve robustness and efficiency. By addressing these realities directly, the work brings ISAC closer to systems that can function reliably outside controlled settings.

Why timing and limits matter

Another common thread across the papers is a focus on constraints that shape real systems. One ICC paper examines sensing-aided communications through the lens of information freshness. In time-critical settings, data that arrives too late can be as limiting as data that never arrives, yet this constraint is frequently sidelined in learning-based designs.

By treating freshness as a design variable, the work exposes trade-offs between accuracy, delay, and resource use. It reflects a wider shift in 6G research, where gains in performance are weighed against latency, overhead, and system complexity.

Another line of work examines dynamic resource allocation in cell-free massive MIMO systems. Cell-free architectures replace traditional cell boundaries with distributed access points that jointly serve users. While this improves coverage and robustness, it also increases coordination complexity.

Learning-based resource allocation offers one way to manage that complexity, particularly in environments where channel conditions and traffic patterns fluctuate. The work explores reinforcement learning as a way to manage coordination when many access points must respond to changing conditions at the same time.

”Networks are evolving to connect data and AI-based applications and agents. This will pose challenges to the networks, e.g., via increased uplink traffic. Machine learning based tools have also opened new venues for technology design and implementation, as seen in many of the papers,” says Markku Juntti, Professor of Wireless Communications and leader of the Wireless Connectivity research area at 6G Flagship.

A pattern across different problems

The papers address different technical questions, but they converge on the same practical problem. How to make learning-based techniques work when wireless systems are no longer stable, isolated, or centrally controlled. Across sensing and communication, beam adaptation, information freshness, and distributed architectures, the emphasis shifts from idealised models toward methods that tolerate uncertainty and change.

Taken together, the papers show researchers returning to the same system-level problems across different technical settings. As sensing, communication, and network architectures become more tightly coupled, the work focuses on how learning-based methods behave when conditions change and coordination becomes harder.

Seen this way, the acceptances capture a moment in which 6G research is becoming less about proposing new techniques and more about understanding how those techniques perform under pressure.

Read featured accepted papers (open access via arXiv)

The full list of authors is available via the conference proceedings and linked papers.

Created 10.2.2026 | Updated 10.2.2026