SoftBank Develops Edge AI System for Autonomous Vehicle Remote Support

Tokyo, Japan—SoftBank Corp announced the development of a new traffic-understanding multimodal AI system that operates on low-latency edge AI servers and aims to automate autonomous vehicle operations fully.

The system, designed to enhance safety and reduce operational costs of autonomous vehicles, started testing at Keio University's Shonan Fujisawa Campus (SFC) in October 2024.

The AI system analyzes front-view camera footage from autonomous vehicles and generates recommended actions for safe navigation by processing multiple data types, including visual and sensor information, in real-time through SoftBank's Multi-access Edge Computing (MEC) infrastructure.

In initial tests at SFC, the system successfully demonstrated its capability to identify potential risks, such as pedestrians potentially hidden by stopped vehicles near crosswalks, and issue appropriate safety commands.

The solution is being tested in autonomous driving trials conducted by MONET Technologies Inc., with remote operators using AI-generated insights to guide vehicles. 

Plans include direct AI-to-vehicle communications to achieve fully unmanned operations.

Hirotane Tanba, Executive Officer of SoftBank's Data Platform Strategy Division, said the successful development and testing of the proprietary traffic-understanding multimodal AI demonstrates how the fusion of telecommunications and AI can address social challenges.