[ITmedia Mobile] Infrastructure strategies for the AI era demonstrated by Docomo, KDDI, Softbank, and Rakuten Mobile “Uplink” and “low latency” will become important in the future

CEATEC 2026 was held for four days from October 13th to 16th at Makuhari Messe in Chiba Prefecture. On the 14th, the second day of the exhibition, executives from four companies gathered: NTT Docomo, KDDI, Softbank, and Rakuten Mobile. The theme was "AI x 6G" and the talk was about how to integrate AI and communications. All four companies are incorporating AI into their networks and using it to improve communication quality and save power, and this was an opportunity to introduce their efforts.

On the other hand, as the use of AI expands and spreads throughout society, the required networks are also changing. When humans communicate using smartphones or smart glasses, upstream bandwidth is required, but in addition to this, low latency is also important for robots and autonomous driving. How are the four companies working on "AI for networks" and "networks for AI"? We have summarized the highlights from the lecture.

What are Docomo, KDDI, and Softbank's efforts to improve wireless quality with AI?

The frontier model of AI is evolving at a faster pace than ever, and carrier networks are also beginning to benefit from and be affected by it. While AI is used to control base stations and entire networks, as AI becomes more pervasive in society, how to manage that traffic is becoming a new challenge for carriers. At CEATEC, each company's AI initiatives were showcased.

In terms of utilizing AI, many companies have already introduced AI to control networks, improve communication efficiency, and even save power, and many have successfully conducted demonstration experiments aimed at deploying it in commercial environments. For example, DoCoMo's demonstration experiment uses AI to optimize the control signals that control wireless terminal connections and the pilot signals within them.

Docomo, in collaboration with Nokia and SK Telecom, eliminated the pilot signal required for channel estimation and instead utilized AI. ``By placing a GPU in the CU (Central Unit) and linking it with devices, we can increase frequency transmission efficiency.'' (Representative Director and Vice President CTO, CAIO Juichi Tsubotani) The experiment was successful. This allows the band used for pilot signals to be used for data transmission and reception, enabling faster communication.

KDDI has introduced distributed AI, which uses AI to allocate control signal resources and eliminates terminals that become unable to communicate, into 100 congested areas across the country and conducted demonstration experiments. Previously, control signals were uniformly assigned, and in congested areas, overflowing terminals became uncontrollable and communications could not begin. This initiative uses AI to disperse this information and distribute it widely and thinly, making it possible to communicate across the board.

Kazuyuki Yoshimura, Executive Officer, Senior Managing CTO, and General Manager of Core Technology Division at KDDI, talks about the benefits of introducing AI, saying, ``In order to improve the situation during congestion, it would be nice to have a huge number of parameters that correspond to each surrounding area, but this was not something that could be done by humans.'' In this demonstration experiment, we succeeded in increasing upstream speeds by 60% during congested times and reducing the number of areas with low speeds by 30%. AI ``judges the settings, which leads to quality improvement'' (ibid.).

SoftBank demonstrated a radar detection function that operates on the AI-RAN-based AITRAS, which is scheduled for commercial deployment. Detects radar systems used by the U.S. Navy from 5G signals in real time and suppresses interference. With AI, ``we are also able to determine subsequent behavior,'' said Hideto Funayoshi, Technology Supervisor, Advanced Wireless Technology Division.