Ready to transition? Strategies to bring drivers back into the loop upon requests to intervene in conditionally automated driving

Abstract

Technology advances and regulatory efforts allow car manufacturers to introduce conditionally automated driving systems into the market. These systems permit drivers to temporarily disengage from the driving task. However, as long as systems are not fully automated, transitions between different levels of automation are necessary and imply shifts in the driver's role and responsibilities. For example, inattention or distraction may impair the driver's ability to take over safely in the case of system limits. Therefore, this work investigates the effect of two transition strategies - monitoring the environment and initiating a maneuver - on bringing drivers back into the loop and ensuring safe driving performance during requests to intervene in Level 3 automated systems (also referred to as take-over requests). Additionally, these strategies could be employed to keep drivers in and on the loop during transitions to lower levels of automation without requiring a manual take-over. For example, when a required driving maneuver would violate traffic rules. In a driving simulator study, 36 participants experienced a conditionally automated drive on a rural road. While participants performed a non-driving-related task (NDRT), the automated driving system approached a rear-end collision blocking the vehicle's lane. In experimental trials, the system requested drivers either to monitor the traffic scene or to decide when to initiate an evasive maneuver. Depending on traffic regulations (overtaking ban present or absent), the system could not solve the situation and issued a request to intervene, or the system could perform the evasive maneuver itself. This resulted in a 2 x 2 within-subjects design with transition strategy (monitoring, initiating) and transition type (request to intervene, no request to intervene) as independent variables. None of the transition strategies improved driving performance, shortened take-over time or affected drivers' gaze behavior. The transition type affected trust in monitoring trials: when monitoring was not followed by a request to intervene, trust in the automated driving system decreased. Results are discussed in light of the suitability of both strategies for different transition contexts.