Distinguished Professor, Department of Electrical Engineering, National Taiwan University Wan-Jiun Liao Building Intelligent, Energy‑Efficient, and Resilient Networks in the 6G era
As global communications transition from 5G toward 6G, the core network design philosophy is undergoing a fundamental paradigm shift. 6G is no longer defined by a mere linear performance improvement in bandwidth, speed, or latency. Instead, it represents an overall evolution toward networks that are intelligent, resilient, and sustainable. Within this framework, key technologies such as AI-native networks, Non-Terrestrial Networks (NTN), semantic communications, and Integrated Sensing and Communications (ISAC) are collectively shaping the development of next-generation communication systems.
In the 5G era, artificial intelligence primarily functioned as an add-on optimization tool, mainly used in resource allocation and traffic prediction. At the core of 6G, however, lies the concept of being AI-native, where AI is embedded directly into the network architecture as the central decision-making engine. Future networks will feature continuous learning and self‑evolution, enabling them to sense environmental changes in real time, acquire and track Channel State Information (CSI) and traffic patterns, and make proactive adjustments. This transformation advances networks from traditional reactive control to proactive control, enabling self-optimization, self-healing, and early risk detection, thereby significantly enhancing system resilience and service reliability.
Building upon this foundation, 6G further equips networks with sensing capabilities. By sharing hardware and spectrum resources, communication and sensing are deeply fused into an Integrated Sening and Communications (ISAC) architecture, enabling simultaneous data transmission and environmental sensing. For instance, Intelligent Transportation Systems (ITS) can track real-time vehicle speed and direction, dynamically adjust beams, and reallocate resources to ensure connection persistence under high‑mobility conditions. In smart manufacturing, multiple robots can collaborate through shared sensory data to maximize both efficiency and safety. This convergence transforms the network from a simple transport layer into an intelligent system with contextual awareness.
AI deployment is also redefining the essence of communication. With the rise of generative AI, reasoning-based AI, and multimodal applications, communication is increasingly shifting toward information exchange led by AI agents. Against this backdrop, semantic communication has emerged as a critical technology. Unlike traditional communication models that prioritize bit-level accuracy, semantic communication focuses on transmitting only the information that is relevant to a specific task. By enabling AI to interpret data content and context, the system can extract the core semantics required for decision-making. This task-oriented communication model significantly reduces data volume and spectrum usage while directly improving energy efficiency.
Apart from the transformation of communication models, the network domain itself is expanding. One of the major developments in 6G is Non‑Terrestrial Networks (NTN), which integrate low earth orbit (LEO) satellites, unmanned aerial platforms, and terrestrial base stations into a Space-Air-Ground Integrated Network (SAGIN). This design not only overcomes the limitations of terrestrial coverage but also supports high‑mobility and wide‑area connectivity. However, the highly dynamic topology introduced by NTN renders traditional connection management and handover mechanisms inadequate. As a result, AI-native mechanisms become the critical foundation, requiring the system to instantly synthesize satellite orbits, user mobility, and spectrum conditions to enable cross‑layer and cross‑domain decision‑making. This is especially critical between LEO satellites and ground terminals, where both sides are in rapid motion and the connection window is extremely short. By employing predictive AI models, the system can anticipate the optimal connection timing and execute path reconfiguration and handovers within milliseconds, achieving near-seamless cross-network connectivity.
Driven by these technological advances, the energy footprint of 6G systems has become a critical concern, making energy-efficient design a paramount priority. Broadly, this can be divided into two approaches: AI for Green and Green for AI. AI for Green utilizes AI to optimize network operations to reduce aggregate energy consumption, such as dynamically shutting down idle resources or adjusting power distribution based on real-time traffic loads. Green for AI focuses on reducing the computational cost of AI itself, including the development of lightweight models and high-efficiency inference architectures. In addition, offloading energy-intensive model training to the cloud or core nodes while deploying ultra-low-latency inference at edge devices is an important strategy for maximizing energy efficiency.
Driven by rising energy and computational demands, extending AI data centers into outer space has emerged as a new frontier. Space AI data centers harness solar energy and the naturally low temperatures of space for passive cooling, thereby reducing the energy burden on terrestrial facilities while integrating seamlessly with the NTN framework. Looking ahead, a space‑ground collaborative computing model may evolve, in which specific model training or inference tasks are executed on satellite nodes and then relayed back to terrestrial systems via high-efficiency communication links. Nevertheless, this architecture faces significant challenges that require further research and validation. These include computing offload, data synchronization, and coordination over high-latency links.
Ultimately, the evolution of 6G is not defined by a single, isolated technological breakthrough but by the highly synergistic advancement of multiple critical technologies. AI-native networks provide intelligent decision-making capabilities; NTN expands connectivity beyond terrestrial limitations; semantic communication reshapes data transmission; ISAC equips networks with environmental sensing; and space AI datacenters redefine computing architecture. Together, these technologies converge on a single, unified objective, which is to build a self-learning, autonomous, and energy-efficient intelligent resilient network. In the future, communication systems will no longer serve merely as information-delivery infrastructure but as the core platform for the operations of a highly intelligent society, striking a masterful balance between extreme performance, unwavering reliability, and long-term sustainability.
Amid these trends, the Hon Hai Research Institute has already established a robust foundation in AI, LEO satellite communications, and electric vehicles. With a distinct advantage in cross-domain integration and system-level deployment, the Institute is well positioned to play a pivotal role in the development and application of key 6G technologies. In doing so, it will make substantive contributions to advancing human society toward a new stage of heightened intelligence and cyber physical convergence.
