Does Adaptive Mode Transition Contribute to Better Driver Intervention in Highly Automated Driving?

Abstract

This paper presents an adaptive mode (level) transition in highly combined driving automation in which the mode of a system could adaptively shift to any level including SAE level 3 (conditional automation, CA) to level 2 (partial automation) based on the driving environment. We show the effects of the adaptive transition on the take over of car control by a human driver and driving behavior after intervention when the system issues a response to intervene. A driving simulator experiment is conducted to collect data during the transition from automated control to manual driving in three scenes: obstacle on a driving lane, blurred lane mark, and stopped car ahead. Results indicate that the interventions of drivers who experience the adaptive transition are delayed in comparison to those who experience only the fixed transition. The adaptive transition is conducive for drivers to stop the car for preventing a potential collision with a stopped car ahead. Owing to the adaptive transition, drivers perceive a critical hazard after taking over car control and provide a rapid response. In addition, during the adaptive transition, drivers prefer verbal messages to the simple “beeping” message.

Effects of Control Transition Strategies and Human-Machine Interface Designs on Driver Performance in Automated Driving Systems

“…This approach termed as the Adaptive Control Transition (ACT) in this study (see Table 1). Few studies have considered ACT as opposed to an immediate handover of control to the driver (21)(22)(23) . We expect ACT might offer the driver additional time to comprehend the situation while ensuring that the ADS continues to maintain safe operation when the transition from L3 to another state, e.g., L0 or L2.…”

“…There have been few studies explicitly comparing FCT to ACT and examining the advantages of ACT over FCT in aspects such as driver performance. Zhou et al (22) explored the effects of an ACT in L3, where the system could shift from L3 to L2 based on the driving environment. The results revealed that drivers were more responsive to critical hazards using ACT and preferred HMI verbal messages over simple beeping sound.…”

“…The scenarios were designed based on the framework proposed by Zhou et al (22) . To mitigate learning effects, the order in which participants completed Trials A, B and C was randomized.…”