Mobile communications systems have become essential infrastructures supporting digital demands from all industry sectors. The current 5G networks are programmable, flexible, and agile, capable of meeting differentiated service level agreement requirements for distinct services and tenants. However, these systems are expected to fall short in meeting the stringent performance requirements of future use cases, such as massive digital twins and applications requiring perceived zero latency.

To address these challenges, the sixth generation (6G) mobile system is expected to provide smart connectivity to a plethora of extremely heterogeneous devices, supporting entirely diverse classes of services with outstanding performance while making the communication infrastructure fully transparent to the applications. The next generation of real-time mission-critical applications requires high resilience and low latency coordinated actions. For example, high-precision robot control and autonomous vehicles cannot afford millisecond latency while factory automation over wireless links demands sub-ms E2E latency.

The new generation of ultra-reliable low-latency communication (URLLC), coined as eXtreme URLLC (xURLLC), requires extra decisions to be made at the network edge, faster and more reliably. The massive amount of data generated at the network edge by autonomous devices cannot be conveyed to the cloud without large delay overhead and high-capacity requirements. Therefore, intelligence needs to be pushed to the network edge, while exhibiting tight coordination among RAN, transport, and computation resources.

The expectation for pervasive network intelligence, setting AI as the cornerstone of mobile communication systems and services, calls for its native integration on 6G. AI and infrastructure programmability have the potential to increase service quality, expand real-time services footprint, and decrease operational costs. Distributed and cooperative AI techniques are envisioned to play a key role in empowering sustainable, fully distributed self-managing capabilities in 6G.

Moreover, 6G systems are expected to provide deep slicing, modular micro-services, and flexible function placement depending on the hardware needs, enabling granular use case instantiation and service level assurance with efficient resource and energy consumption. Pursuing deep programmability of the network in 6G both vertically and horizontally can provide significant benefits, including dynamic traffic engineering at wire-speed, slicing and multi-tenancy, and offloading network functions to the data plane.

In conclusion, embracing AI in 6G is expected to have a significant influence on the corresponding network architecture design, and deserves a clean-slate rethink. These proposed innovations will create a significant impact on the wireless communication market, leading to new applications and services that will improve people’s lives.