A research team from the Hong Kong University of Science and Technology (HKUST) proposed a new Reasoning-Empowered Task-Oriented Communication (TOC) framework that enables AI agents to determine what information to share, when communication is necessary, and why a communicate should occur. The research could pave the way for collective intelligence in future 7G wireless networks.
Led by Prof. Khaled B. LETAIEF, New Bright Professor of Engineering and Chair Professor of Department of Electronic and Computer Engineering (ECE) at HKUST, the team includes XIE Songjie and LI Hongru, both PhD candidates in HKUST's ECE; Dr. WANG Zixin, Research Assistant Professor of ECE; Prof. SONG Shenghui, Associate Professor of ECE and the Division of Integrative Systems and Design; and Prof. ZHANG Jun, Professor of ECE. The research, titled "Towards Reasoning-Empowered Task-Oriented Communication for Agent Networks", was published in the prestigious academic journal npj Wireless Technology.
This research presents a roadmap for how communication and reasoning may converge in the coming decades, helping shape the evolution of wireless networks beyond 6G. It offers an early vision for the development of future 7G wireless systems capable of supporting large populations of autonomous AI agents capable that can reason, collaborate and make collective decisions.
Today's communication systems are engineered to deliver data as accurately and reliably as possible. However, future smart cities, autonomous vehicles, healthcare platforms and industrial systems may involve millions of intelligent agents simultaneously exchanging information. Simply scaling today's communication architectures may not be enough. Without new approaches, communication itself could become a bottleneck that limits the collective intelligence of future AI ecosystems.
To address this challenge, the HKUST team's Reasoning-Empowered TOC framework integrates reasoning into the communication process. Before transmitting, an AI agent would evaluate the value of an exchange, identify the most relevant information, choose the right recipients and anticipate how the message could affect future decisions.
The proposed cognitive loop has three capabilities. First, intent interpretation converts a high-level human or machine objective, such as "keep a video call stable", into a structured communication goal. Second, automated formulation and optimization selects an appropriate communication strategy while balancing bandwidth, power, latency and robustness as conditions change. Third, proactive foresight uses a world model, an internal representation of relevant conditions, to anticipate changes in the environment, user mobility and task requirements before performance deteriorates. Together, cognition guides communication, and communication strengthens collective cognition.
Potential applications could include vehicles and roadside systems exchanging only the insights needed to avoid hazards, clinical agentsprioritizing signals for time-sensitive decisions, and machines and digital twins coordinating maintenance before a fault interrupts production.
Prof. Letaief said, "Tomorrow's wireless networks will not simply connect devices. They will connect intelligence. The next transformative leap will be connecting intelligent agents capable of reasoning, planning and autonomous collaboration. Future networks will not simply transport information; they will enable collective intelligence. Reasoning-empowered task-oriented communication offers a pathway toward wireless systems that can determine what information matters, when communication is needed and why it should occur."
He added that this research is a compass for 7G research, not a finished technical standard. It maps the open questions that must be solved to make the vision real, including new theoretical foundations, scalable multi-agent coordination, the stability of communication-reasoning loops, trustworthy AI decision-making, and common standards and benchmarks.
