Aira develops AI-based channel estimation and prediction xApp

Aira Technologies, a pioneer in the application of Machine Learning (ML) to radically improve wireless telecommunications, announced that it has been awarded a Requirements Compliant Bronze Badge from the Telecom Infrastructure Project (TIP), for delivering an AI-based High Fidelity MIMO Channel Estimation and Prediction application (x2 xApp). The requirements of the x2 xApp have been specified by the RAN Intelligence & Automation (RIA) subgroup of the TIP Open RAN 5G NR Project Group. The TIP RIA subgroup is led by Vodafone, BT and T-Mobile USA.

The TIP RIA subgroup is focused on bringing the benefits of Artificial Intelligence (AI), Machine Learning (ML), and data science technologies to the Open RAN ecosystem. In particular, AI/ML algorithms running as applications (xApps and rApps) on top of the RAN Intelligent Controller (RIC) platform can help operators improve network utilization and customer experience through automation. The algorithms themselves are trained with actual network data and infer how to manage network resources optimally. In this way, Mobile Network Operators (MNOs) are able to contain operational expenses, even as they scale their networks, leading to more competitive Total Cost of Ownership (TCO). The RIA subgroup has developed AI-based use cases for Radio Resource Management (RRM) and Self Organizing Network (SON) that will help improve network coverage, capacity, handover and interference in an automated manner. Other AI-based use cases address the optimization of Massive MIMO system performance, thus increasing their spectral efficiency.

“The application of ML to wireless baseband processing is an industry first. The RAN Intelligence and Automation (RIA) working group of TIP has articulated an outstanding need for a near real-time application to perform better channel estimation to improve MU-MIMO performance,” said Anand Chandrasekher, co-Founder and CEO of Aira Technologies. “Aira’s xApp delivers against this need by providing high-fidelity beam management through improved channel estimation and prediction. We are proud of the recognition from the TIP organization and very pleased to be working with our TIP partners.”

“Radio Access Networks (RAN) operate in varied network conditions and radio environments. ML presents a natural framework to classify these conditions accurately, and process the RAN signals optimally for each of these conditions,” said Ravikiran Gopalan, co-Founder and CTO of Aira Technologies. “We are seeing tremendous MU-MIMO throughput gains from our ML-based channel prediction xApp, and we are working on applying our ML framework to other RAN functionalities.”

The setup at the TIP community lab sponsored by Meta in Menlo Park, California, consists of a Viavi UE emulator and Viavi channel emulator that simulates varying channel conditions as seen by a Foxconn (4T4R) RU. The RU itself is connected in an end-to-end fashion to the Aira DU, a CapGemini CU and a CapGemini Core. Aira has defined a service model that passes data from the DU to the VMware RIC via the E2 interface. The Aira AI-based High Fidelity MIMO Channel Estimation and Prediction xApp runs atop the RIC. This xApp estimates and predicts the channel which is then compared to ground truth as established by the Viavi channel emulator. The output of the xAPP is the prediction of channel evolution in-between S-slots, which results in improvements in RAN throughput and coverage.

“The TIP OpenRAN project group is focused on tools that bring in automation through machine intelligence to enhance RAN management and performance. TIP has worked with leading MNOs and vendors to define differentiated use cases, and demonstrated them on RIC platforms over the past two years. One of the focus areas for TIP RIA are near-RT RIC applications that focus on improving MU-MIMO and maMIMO performance,” said David Hutton, Chief Engineer at TIP. “We are very excited to be working with our participants to unlock additional use cases for near-RT control and optimization of OpenRAN networks.”

CT Bureau