On October 22, KDDI announced that it has developed an AI agent
to support “1-to-N remote monitoring” for Level 4 autonomous driving,
in which one person watches over many vehicles.
The agent analyzes video and other information from vehicles
where an anomaly has occurred,
and presents response priorities and recommended actions to the monitoring operator.
KDDI will begin trials in November and aims for early
real-world deployment in 2027 or later.
Background: When Several Vehicles Stop at Once, Where Do You Start?
According to KDDI, Level 4 autonomous driving with remote monitoring
requires a remote monitoring system
and a designated autonomous operation supervisor (the monitoring operator),
who is responsible for checking vehicle status and responding to anomalies.
The company says a setup in which one person monitors many vehicles
is important for addressing labor shortages while keeping operating costs down.
KDDI’s technical explainer gives the example of sudden heavy rain
causing vehicles in the same area to stop one after another.
Each vehicle stops in a different place,
and after stopping, fallen trees or flooded roads may occur.
The burden on the monitoring operator of continually switching between
multiple video feeds to check them is a challenge.
How It Works: Analyzing Changes After a Stop and Updating Priorities
Triggered by an anomaly alert, the AI obtains
the location information, camera video, and other data of the vehicle concerned.
It analyzes data related to the alert,
and rather than stopping at the initial assessment,
updates response priorities as surrounding conditions change.
The technical explainer gives an example in which priority is set low
when a vehicle has stopped safely on the road shoulder,
and raised to high if a fallen tree or flooding later appears ahead
and additional response becomes necessary.
The agent continues to track changes
even while the monitoring operator is handling another vehicle.
What Changes: Reviewing Recommended Actions Together with the Reasons
A VLM (vision-language model) is used to analyze
vehicle camera video and various other information.
In addition to recommended actions, such as making emergency calls
or arranging for staff to be dispatched to the site,
the reasons for the assessment are displayed on the monitoring screen
in images and text.
The monitoring operator checks the key points and makes the final decision.
The reasons are shown so that the monitoring operator can focus on checking
the data underlying the recommended action.
KDDI believes this lets operators quickly confirm
whether the presented information is appropriate,
leading to faster and more accurate final decisions.
Trials from November: Testing Based on Operational Experience
In July 2025, KDDI successfully conducted a trial
on the Shin-Tomei Expressway combining anomaly detection,
dispatch of staff to the site, and moving vehicles to safety
for autonomous trucks.
Since October 2026, it has also been remotely monitoring
autonomous route buses in Tsukuba.
The trial of the new AI will be conducted at KDDI’s Mobility Control Center.
It will simulate situations in which anomalies occur simultaneously
in many vehicles and in which surrounding conditions change after vehicles stop,
to verify effectiveness,
and KDDI will work with KDDI Smart Mobility to make improvements.
At this point, the company has not stated how many vehicles can be monitored
or how much the workload will be reduced.
LC Signal’s Perspective: Conditions Keep Changing Even After a Vehicle Stops
A key feature is that the system does not just report an anomaly when a vehicle stops,
but follows subsequent changes and updates the order of response.
A mechanism that keeps tracking conditions
while a person is handling another vehicle
seems likely to be useful when multiple problems occur at the same time.
In the trials, we will watch how far the presentation of priorities and reasons
can support the monitoring operator’s response.
Key Sources
-
KDDI official announcement (Japanese), October 22, 2026:
レベル4自動運転車両の1対N型遠隔監視を支援するAIエージェントを開発
(AI agent developed to support 1-to-N remote monitoring
of Level 4 autonomous vehicles) -
KDDI official technical explainer (Japanese), October 22, 2026:
AIエージェントで支える自動運転の遠隔監視
(Supporting remote monitoring of autonomous driving with AI agents)
This article was prepared with AI assistance using publicly available information
as of October 22, 2026.
Announcement dates follow the dates shown on the official pages.
Developed functions are distinguished from plans for future trials
and real-world deployment.
“LC Signal’s Perspective” reflects this publication’s interpretation
of the available information.