Next Decade of Telecommunications Artificial Intelligence

Author(s): Ye Ouyang, Lilei Wang, Aidong Yang, Tongqing Gao, Leping Wei, Yaqin Zhang

Paper URL: https://www.sciopen.com/article/10.26599/AIR.2022.9150003

Pages: 23

Summary: This paper contains a comprehensive review and forward-looking perspective on the integration of mobile communication and AI from 5G and beyond 5G (B5G).

Sections Summary

1. Mobile Communication and AI

• Mobile communication has evolved from analog to digital, from voice-only to voice and data, and from circuit-switched to IP-based systems.
• When 4G was established, the industry started to emphasize automation and intelligence to handle increasingly complex communication networks and personalized services.
• We are now generating vast amounts of data which facilitates the application of AI in the communication field.
• The vast data generated today provides a foundation for applying AI in telecommunications.

2. Development Roadmap of Mobile Communication and AI

• The mobile industry evolved through 3GPP standards from 2G to 5G, with AI concepts introduced in 2008 via self-organizing networks (SON).
• Telecom AI initially saw limited use but grew rapidly post-2017 with 3GPP’s research on 5G features, network optimization, and data analytics.
• Organizations like 3GPP, ETSI, O-RAN, and ITU have developed AI frameworks and standards, including federated learning integration into 5G.

3. Development of Telecommunications AI

• Traditional communication systems rely on mathematical models for signal processing tasks like modulation, error correction, and noise filtering. Deep learning (DL) bypasses the need for explicit models, instead learning patterns directly from data.
• Communication systems are hierarchically structured, similar to IT microservices. AI offers potential for holistic system optimization rather than isolated subsystem improvements.
• AI in Network Infrastructure (base stations, routers, terminals...):
  • Wireless Access Network: AI techniques (e.g., DNNs and CNNs) optimize tasks like channel quality, signal detection, and resource management. SON aims for self-configuration and optimization but faces commercialization challenges due to vendor-specific implementations.
  • Core Network: NWDAF (Network Data Analytics Function), standardized in 2017, integrates AI into the 5G core for mobility, QoS, and network element management. O-RAN's RIC enables AI-driven RAN management, though still in early trials.
  • Transport Network: AI aids in monitoring, optimizing, and ensuring service quality but is still in early research stages.
  • Terminal: Terminals report performance data for network optimization using AI in SON and OSS systems. Focus is also on integrating intelligence within terminals and chips.
• AI in Network Management (how networks are maintained, optimized, secured):
  • MDAF: Analyzes network data to optimize performance and support QoS, standardized by 3GPP.
  • ENI Engine: Adapts networks over time based on feedback to improve reliability, defined by ETSI.
  • Network OSS: Uses predictive analytics for automated operations, fault handling, and resource optimization.
• Cross-domain AI integration enhances telecom business applications, including customer service, billing, policy control, and private networks.

4. Next Decade of Telecommunications AI

• AI in Telecom Network Infrastructure:
  • Wireless Access Network: 3GPP SA5 and RAN3 are advancing SON (Self-Organizing Networks), LTE, and NR data utilization, with RAN-DAF (Radio Access Network Data Analytics Function) for analytics and decision-making.
  • Core Network: NWDAF supports real-time data analysis for network slicing, load balancing, and integration with multi-access edge computing (MEC) for industry applications like healthcare.
  • Transport Network: Distributed architectures leverage blockchain for secure elastic computing, transitioning from IPv6 to IPv6+ for business-specific needs.
  • Terminal: AI-powered terminals enable seamless network switching, intelligent beamforming with deep reinforcement learning, and improved wireless access.
• AI in Network Management:
  • MDAF enhancements by 3GPP SA5 for better integration with network functions.
  • ETSI ENI to define interfaces for deployment strategies and data processing.
  • IBN (Intent-Based Network) simplifies management in multi-vendor settings, with active research by China Telecom.
  • AI-driven O&M systems link 3GPP SON, NWDAF, O-RAN RIC, ETSI ENI, and 5G OSS to support orchestration and monitoring.
• AI will enhance BSS intelligence, finance, and operations for customized customer experiences and new service models.
• Cross-domain intelligence evolves from SLA-based to ELA-based customer experience systems.
• Private network applications anticipate intelligent vehicles, telemedicine, and smart cities.

5. Expectable Future: Comprehensive Promotion of Telecommunication Intelligentization in the Next Decade

• 6G will integrate air, sea, and land systems, using AI to solve complex optimization challenges across ecosystems.
• ITU-R’s formal 6G standard is anticipated around 2027–2028.
• SON and O-RAN RIC systems will advance for AI-enhanced wireless optimization.
• Federated learning, blockchain, and privacy-preserving computing will ensure data security.
• Early commercial 5G private networks will leverage AI for SLA guarantees and resource optimization, progressing toward real-time QoS monitoring in hybrid and standalone networks.
• Digital twin technology, network simulation, and AI will enable advanced lifecycle management (LCM).

Questions/Discussion Points

• AI and 5G are considered General Purpose Technologies (GPTs). GPTs are foundational technologies that have broad applications and potential for significant impact on the economy and society, characterized by their ability to spur innovaction, drive productivity and transform multiple industries.
• This whitepaper was written in 2021, it's now 2024 Q4 - how far did we come with the progress now..?