Verification of Continuous Recovery Operation with Teleoperation System for Autonomous Vehicles

Abstract

Remote-type autonomous driving is being introduced as a mechanism to ensure the imperfection of autonomous driving systems, since these systems can only run automatically under pre-designed conditions. In remote-type autonomous driving, a human operator at a distance from an autonomous vehicle monitors the vehicle status using a teleoperation system and operates the vehicle. In this study, we target the situation in which multiple autonomous vehicles cannot continue to drive themselves at the same time and a single operator is required to operate them continuously to recover from the situation. The experiment is conducted to verify human error and work efficiency issues with a driving simulator. In addition, the influence of the characteristics of the vehicle situation on the restoration operation is investigated. As a result, although no issues related to human error are identified, it is suggested that road surface markings may have an impact on work efficiency in these operations. Furthermore, we confirm that the difficulty of recovery operations tends to increase when the factors that the autonomous driving system stops are not displayed as images from the onboard camera.