Impact of non-driving related task types, request modalities, and automation on driver takeover: A meta-analysis

The transition towards fully automated driving necessitates human intervention in specific scenarios, making it crucial to understand the factors influencing driver takeover performance. This meta-analysis systematically reviews 37 studies selected from an initial pool of 1945, focusing on the impact of non-driving related task (NDRT) types, takeover request (TOR) modalities, and levels of automated driving (LAD) on driver response and vehicle control during takeover events. The findings reveal that engagement in multiple NDRTs significantly delays driver response times and degrades control over vehicle dynamics, particularly in critical lateral and longitudinal maneuvers. Furthermore, multimodal TORs are more effective in eliciting timely and accurate driver responses compared to unimodal TORs, which often result in suboptimal performance. Additionally, manual driving (L0) improves emergency response but comes with a higher driving workload compared to conditional automated driving (L3) takeovers. These insights underscore the need for optimized TOR strategies and the development of advanced multimodal systems to enhance driver readiness and safety in automated driving environments.