How can 3GPP leverage AI and edge computing to optimize IoT performance and efficiency in 5G networks?

1 AI for IoT

AI can help 5G networks to manage the massive amount of data and devices that IoT applications generate and consume. AI can enable network automation, self-optimization, and adaptive resource allocation, reducing human intervention and operational costs. AI can also enhance the quality of service, security, and privacy of IoT applications, by detecting and mitigating anomalies, attacks, and faults. For example, AI can help 3GPP to implement the concept of network slicing, which allows different IoT applications to have customized network resources and performance levels.

2 Edge computing for IoT

Edge computing is the paradigm of processing data closer to the sources and destinations, rather than in centralized cloud servers. Edge computing can reduce the latency, bandwidth, and energy consumption of IoT applications, by offloading computation and storage tasks from the network core to the network edge. Edge computing can also improve the reliability, security, and privacy of IoT applications, by minimizing the exposure and dependency on the network infrastructure. For example, edge computing can help 3GPP to support the ultra-reliable low-latency communication (URLLC) service class, which requires sub-millisecond latency and high availability for critical IoT applications.

3 3GPP standards for AI and edge computing

3GPP has been developing standards and specifications for AI and edge computing in 5G networks, to enable interoperability, compatibility, and innovation. TS 23.501 outlines the overall architecture and functions of the system, including the network functions and interfaces that support AI and edge computing capabilities. TS 23.791 examines the mechanisms and scenarios for distributing traffic across multiple access networks, such as cellular, Wi-Fi, and satellite. TR 23.793 looks into the enhancements of the NWDAF, which is the network function that collects, analyzes, and provides network data for AI applications and services in 5G networks. Lastly, TS 23.735 studies the architecture enhancements and requirements for supporting edge computing in 5G networks, such as edge application enablement, edge discovery, and edge mobility.

4 Challenges and opportunities for AI and edge computing

Despite the potential benefits of AI and edge computing for IoT in 5G networks, 3GPP must also address and explore a variety of challenges and opportunities. For instance, they must ensure trustworthiness, explainability, and accountability of AI applications and services in 5G networks, particularly for sensitive and critical IoT domains. Additionally, it is important to balance the trade-offs between centralization and decentralization, coordination and autonomy, and privacy and utility of data and computation in AI and edge computing scenarios. Furthermore, fostering collaboration and innovation among different stakeholders such as network operators, service providers, device manufacturers, application developers, and end users in the AI and edge computing ecosystem is essential. Finally, leveraging the synergies between AI and edge computing is key; for example, using AI to optimize edge computing resources or using edge computing to accelerate AI inference and training.

5 Future directions for AI and edge computing

AI and edge computing are rapidly advancing fields that will continue to influence 5G networks and IoT applications. 3GPP must stay up-to-date with the latest developments and trends in AI and edge computing to meet the future needs of the IoT market. Possible future directions include federated learning, neuromorphic computing, multi-access edge computing (MEC), and edge intelligence. Federated learning is a distributed learning approach that allows multiple edge devices to collaboratively train a shared AI model without exchanging raw data, preserving privacy and reducing communication overhead. Neuromorphic computing is a computing paradigm that mimics the structure and function of biological neural networks, offering high performance, low power consumption, and adaptive learning for AI applications. MEC is a framework that enables the integration and coordination of multiple edge computing platforms and resources across different access networks and domains. Finally, edge intelligence combines AI and edge computing to enable intelligent decision making, coordination, and adaptation at the edge of the network, empowering IoT devices and applications with autonomy and agency.

6 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?