Japan-based mobile operator SoftBank improved its radio access design in the Tokai region by using a network automation service from Ericsson.
Ericsson's elastic radio access (RAN) design uses machine learning and big data analytics to enable automation. The end result is an improved user experience and a reduction in lead time of 40% over traditional network design methods.
While SoftBank is seeing improved benefits now with the new design, which it will use in other geographical areas, automation and the use of artificial intelligence will have a big impact in the near future on the rollout of 5G technologies and designs.
"I would say that this is a solid incremental step toward planning and operating much denser radio networks, which 5G will certainly need," said IDC's Andy Hicks, research director, EMEA telecommunications and networking. "The AI aspect is relevant because it’s inevitable in the long term as network complexity grows beyond the point where we can manage these aspects manually."
SoftBank used Ericsson's service in dense, urban clusters where there was a lot of multiband complexity. Ericsson's service conducts an analysis of the radio network environment, which included taking the cell coverage overlap, signal strength and receive diversity into consideration.
The large number of possible relations between the cells required a substantial amount of computational power and state-of-the-art machine learning techniques, according to Ericsson. Ericsson was able to cut through these complexities by using its patented network graph machine-learning algorithm.
"There is a huge potential for machine learning in the telecom industry and we have made significant investments in this technology," said Peter Laurin, head of managed services, in a prepared statement. "It is very exciting to see that the new methods have been successfully applied in SoftBank's network. There is a strong demand for this type of solutions and deployments of this service to other tier-one operators in other regions are ongoing."
The service grouped cells in clusters and mined statistics from cell overlapping and looked at the potential for carrier aggregation between cells, which improved performance and lowered opex.
Using these various methods, SoftBank was able to automate the process for its RAN design. Big data analytics were applied to a cluster of 2,000 radio cells, and that data was analyzed to find the optimal configuration.
Ericsson said SoftBank was the first Tier 1 operator to use the AI assisted RAN network design. It took Ericsson six months to develop prior to being delivered to SoftBank.
Ericsson's Network Design and Optimization Artificial Intelligence Accelerator Lab, hosted in Japan and Sweden, is looking to industrialize five AI uses this year for network design and optimization.
Editor's Note: This article was updated with additional information in the last two paragraphs.