Japan-based mobile operator SoftBank, looking to improve radio access network (RAN) design as the process becomes more complex with the move to 5G, tested a network automation service from Ericsson which uses machine intelligence and big data analytics.

The operator applied the service to dense urban clusters with multi-band complexity in the Tokai region. Ryo Manda, radio technology section manager at SoftBank, said the outcomes exceeded its expectations and it is implementing the design method in other areas.

Ericsson said in a statement the foundation of the method is a thorough analysis of the actual radio network environment, for example taking cell coverage overlap, signal strength and receive diversity into consideration. The high number of possible relations between cells requires substantial computational power and state-of-the-art machine learning techniques.

The design method is based on a network graph machine learning algorithm, which Ericsson has patented, the vendor said. The service groups cells in clusters and takes statistics from cell overlapping and the potential to use carrier aggregation between cells into account to reduce opex and improve performance. Compared with traditional network design methods, it cut the lead time by 40 per cent.

With the launch of 5G services, operators’ network configurations will become increasingly complex as cell site density will increase significantly.

Peter Laurin, head of Managed Services at Ericsson, said: “There is a huge potential for machine learning in the telecoms industry and we have made significant investments in this technology. It is very exciting to see that the new methods have been successfully applied in SoftBank’s network.”