by Carlos Ubeda, RIA co-lead, Vodafone and Richard Mackenzie, RIA co-lead, BT

OpenRAN Begins Work on AI/ML Applications for Radio Management

Jul 28, 2020

Today we are announcing the launch of a new subgroup within the TIP OpenRAN 5G NR Project Group – RAN Intelligence & Automation (RIA). The RIA subgroup will enable the OpenRAN ecosystem to leverage Artificial Intelligence (AI), Machine Learning (ML), and data science technologies to better address the RAN market needs and improve network economics. Vodafone, BT, and T-Mobile USA will lead the RIA subgroup.

OpenRAN can help wireless network operators improve customer experience, by expanding capacity and coverage efficiently. These benefits can be achieved by faster innovation cycles, greater supply chain diversity, and automation of network operations that also lower overall network TCO.

TIP’s OpenRAN and OpenRAN 5G NR Project Groups are working to create and deploy open, disaggregated and standards-based radio access networks that will enable attractive use cases for connectivity that benefit society and create economic value. TIP’s OpenRAN momentum is driven by operator trials and deployments, including Vodafone’s field trials in UK, Ireland, Turkey, the Democratic Republic of the Congo and Mozambique, the field trials conducted by Indosat Ooredoo and Smartfren in Indonesia, by Edotco and Celcom Axiata in Malaysia and Etisalat in several markets.


Data science and AI/ML can enhance OpenRAN solutions

OpenRAN solutions can apply data science leveraging AI/ML technologies to significantly improve performance, and operation automation, through algorithms that learn from experience and continuously improve the system. Many RAN features, which traditionally have been manually programmed, will rely on AI capabilities to handle the increasing complexity of current and future networks.

O-RAN ALLIANCE specifications provide a framework to use AI/ML to optimize how radio resources are managed in a 5G network. The framework provides open interfaces and disaggregates the near-RT functions from CU and DU which also simplifies the hardware requirements at the cell site. This is done through applications (xApps for near-RT RIC) hosted on “Radio Intelligent Controller” (RIC) platform which can be implemented for near real-time control (near-RT RIC, connected to the CU through the E2 open interface) and for non-real-time control (non-RT RIC, connected to the near-RT RIC through the A1 open interface).

AI/ML technologies can be used to implement Self-Organizing Networks (SON) and Radio Resource Management (RRM) solutions that improve network coverage, capacity, handover and interference in an automated manner. They could also be applied to optimize the performance of Massive MIMO systems, and thus increase their spectral efficiency. Further, AI/ML can improve user experience with VoLTE/Video quality optimization, Terminal Anomalies Detection and other QoS/QoE based use cases.

AI/ML algorithms, running as applications, on top of the RIC platforms can help operators improve network utilization and customer experience through automation. These algorithms are trained with actual network data and then infer how to manage network resources optimally. In this way, operators will be able to scale their networks without increasing network operation expenses at the same rate.


A new RIA Subgroup within the OpenRAN 5G NR Project Group (PG)

Within the OpenRAN 5G NR Project Group, we are now creating a new RIA Subgroup that will include operators and RAN ecosystem participants to work together on specific use cases where AI/ML and Data Science technology can add value to the RAN. Initial operator participants include BT, Dish, Deutsche Telekom, T-Mobile USA and Vodafone. The RIA subgroup will produce requirements for the use cases, and will engage the OpenRAN ecosystem to develop these use cases and build reference solutions validated in the labs and in field trials. The RIA subgroup will initially target use cases for Radio Resource Management (RRM), Self Organizing Network (SON) and Massive MIMO.

Members of the RIA subgroup will participate in use case testing in TIP Community Labs and in operator trials and will share results with the OpenRAN 5G NR PG community.

“The increasing interest in the industry to embrace OpenRAN specifications and virtual RAN technology creates tremendous opportunities to accelerate service innovation and transform RAN economics through automation. On the other hand, these new approaches also introduce multi-vendor complexity and systems integration challenges that must be addressed to increase trust and demonstrate feasibility. In this broad context, the creation of the RIA working group is timely and relevant, providing a forum for ecosystem collaboration and demonstration of early automation use cases using ML/AI technology” said Anil Rao, Principal Analyst at Analysys Mason.


A collaborative effort across industry bodies

The key objective for the RIA subgroup is to develop and deploy AI/ML based applications for RAN. Towards this, and enabled by TIP’s liaison agreement with O-RAN ALLIANCE, the subgroup will work with open interface specifications from O-RAN ALLIANCE. Further, taking advantage of TIP’s liaison agreement with ONF, the RIA subgroup can leverage open source RIC from ONF’s SD-RAN project. Results of the work from the subgroup could also be adopted by members to make joint contributions to the O-RAN ALLIANCE architecture and interface specifications, to incorporate potential learnings from the labs and trials.

We look forward to the OpenRAN ecosystem actively collaborating in the RIA subgroup on use cases that will leverage the strength of Data Science and AI/ML technologies and open, standard interfaces. Through this, it will be possible for OpenRAN solutions to set new industry benchmarks on performance, efficiency and total cost of ownership.


To learn more and to join the OpenRAN 5G NR project group and the RIA subgroup, visit the project group website here.