When there is no rush to take over: evaluating multi-stage takeover request designs in conditionally automated driving

Conditionally automated vehicles (AVs) allow drivers to divert their attention from the road under specific conditions defined by the Operational Design Domain (ODD) but require drivers to resume manual control when approaching the ODD boundary. In such scheduled takeover scenarios, drivers are afforded more time to restore situation awareness and manage the takeover task at their own pace. This study focuses on a typical ODD exit scenario—freeway exits—and investigates the effectiveness of a multistage takeover request (ToR) design comprising three stages: information, warning, and command. A driving simulator experiment involving 32 participants evaluated how ToR presence (single-stage vs. multistage) and modality (semantic speech, dynamic visuals, or both) affected the takeover process, including takeover strategy, pre-takeover situation awareness, post-takeover vehicle control performance, and subjective evaluations. Results showed that multistage ToR designs significantly improved behavioral performance, including faster eyes-on-road and takeover responses, shorter preparation times, and enhanced perceptual and cognitive preparedness. These designs also enhanced user experience, particularly by increasing trust and reducing mental workload. The performance benefits were most pronounced for female drivers and for those who delayed initiating their takeover response until the command stage. These findings highlight the value of incorporating structured, multimodal ToR designs for scheduled takeover scenarios, and support the development of adaptive systems tailored to individual differences in driver response tyle and characteristics.