Nine parties began testing L4S and network slicing on autonomous-bus monitoring video

Nine organizations began a technical demonstration in Shiojiri, Nagano Prefecture, on November 2, transmitting remote-monitoring video from autonomous buses over a mobile network using L4S together with network slicing. The announcement was made on October 29; the demonstration itself starts today.

The participants are KDDI, Alpico Kotsu, Aisan Technology, the Shiojiri City Promotion Public Corporation, KDDI Smart Mobility, KDDI Technology, KDDI xG Networks, A-Drive and Persol Business Process Design, with TIER IV, Mitsubishi Electric and Alpico Taxi named among contracted and cooperating organizations.

KDDI said network slicing logically separates the monitoring traffic from ordinary subscriber traffic, so the video keeps the quality it needs when nearby smartphone use congests a cell, and that an L4S function deployed on Ericsson network equipment detects the onset and clearing of congestion and dynamically optimizes the video quality in response. The company described the arrangement as a first in Japan for applying L4S congestion detection and video-quality optimization in an autonomous-driving business, qualifying that as its own subsidiary’s finding as of October 29.

Three things are being measured: how far dynamic optimization reduces freezes and severe disruption in the monitoring feed, how much the monitoring traffic affects ordinary subscribers nearby, and whether steadier video reduces the psychological load on a supervisor watching several vehicles. The buses are TIER IV Minibus autonomous electric vehicles. One route runs at Level 4 with a specified-autonomous-operation supervisor in the driver’s seat who performs only the start and end operations and does not drive; a second route runs at Level 2. Public rides open on the first route from November 9 and on the second from December 7. The work sits under a Ministry of Internal Affairs and Communications program, and the stated goal is the connectivity a one-to-many remote-monitoring model would need.