KDDI builds AI agent for 1対N remote monitoring of Level 4 cars

KDDI's move sits inside a regulatory setup: remote-monitoring Level 4 services must have a remote monitoring system and a 特定自動運行主任者 who checks vehicle status and handles anomalies. That requirement makes the supervisor a bottleneck as fleets grow, which is why KDDI frames 1対N型遠隔監視 as important for expanding autonomous vehicle deployment while reducing operating costs under labor shortages.

The hard part KDDI targets is not spotting a single anomaly but handling several at once. When multiple vehicles stop, a supervisor has to grasp each vehicle's condition and surroundings in real time and decide quickly which one to handle first. The agent keeps analyzing conditions after a stop — such as debris flying in or surrounding traffic changing — and updates priority even while the supervisor is busy with another vehicle, using a VLM (vision-language model) to combine camera images with vehicle data and show the recommended action and its reasoning.

KDDI and KDDI Smart Mobility already support municipalities and transport operators from introduction through operation of autonomous and on-demand transport, including remote monitoring via the モビリティコントロールセンター. The November demonstration there, simulating simultaneous anomalies and post-stop changes, is positioned as the validation step before KDDI aims for early social implementation from 2027 onward.