Towards artificial intelligence enabled 6G: State of the art, challenges, and opportunities

With the rapid progress of 3GPP 5G phase 2 standardization and forthcoming commercial deployment of 5G system [1], many re- searchers [2],[3],[4],[5] from academia and industry have started envisioning what 6G will be. Giordani et al.[2] illustrated potential 6G use cases and potential enabling technologies. Zhao et al.[3] surveyed potential 6G technologies emphasizing artificial intelligence (AI), intelligent surfaces, and Terahertz communications. Huang et al.[4] presented a survey of potential technologies that are helpful to realizing sustainable 6G networks. Wikstrom et al.[5]described a set of potential challenges and possible technical components of the future 6G system. 6G is expected to support numerous novel Internet of Everything (IoE) applications such as extended reality (XR), holographic communica- tions, smart environments, and brain-computer interactions. To fulfill the challenging and diverse demands in terms of performance, security and cost, emerging technologies such as ultra-massive Multiple-input multiple-output (MIMO) [6], holographic beamforming, large intelligent surfaces (LIS)[7], Terahertz (THz) communications [8], visible-light communications (VLC), and new network paradigm such as intelligent radio, AI native architecture, space-aerial-terrestrial-ocean integrated 3D networks are under extensive studies. The large dimen- sion, super heterogeneous, extremely complex 6G networks should be flexible enough to provide customized services in dynamic environ- ments. However, it is very challenging to manage the complex system efficiently by conventional manual centric approaches. Planning and optimizing the space-air-ground integrated networks with different types of radio access technologies are facing many unprecedented challenges [9]. The growing diversity of network services and the complexity of network systems make conventional network managing and optimizing approaches relying on mathematical models no longer adequate[10]. All the above-mentioned issues call for new approaches to achieve a 6G system that is flexible, adaptive, and intelligent to support diverse services efficiently without heavily relying on conven- tional management and optimization paradigms.