Temporal Augmentation for Remote AV Takeover: A Dual-Phase Evaluation of Replay-Buffered Interfaces

Remote teleoperation is critical for handling edge cases in autonomous vehicle operations, yet current Takeover Request interfaces often lack retrospective context, impairing operators' ability to rapidly build situation awareness. We developed two graphical user interfaces, a live-view "Actual Interface" and a "Buffered Interface" augmented with a scrub-able replay timeline, and evaluated them through a dual-phase methodology combining simulation-based saliency analysis with a within-subject user study and post-task interviews. Saliency classifiers confirmed that the interfaces are structurally distinguishable in their bottom-up attentional properties. However, Bayesian repeated-measures analysis revealed that scenario complexity, rather than interface condition, dominated variance in cognitive workload, situation awareness, and takeover time. Qualitative data clarified this pattern: participants rarely engaged with the timeline, perceiving the attentional cost of reviewing past frames as prohibitive under time pressure. Our findings advocate for multi-layer evaluation combining computational screening with human-in-the-loop testing and qualitative inquiry to guide Takeover Request interface design.