Bundled or Separated? How ADAS Architecture Shapes Mode Confusion and Driver Control Behavior

Effective human–automation interaction requires alignment between driver mental models and system control logic. In driver assistance systems, this alignment is shaped by architecture: how functions are bundled and states are communicated. This study examined how three ADAS architectures influence driver understanding of system state, responsibility allocation, and driving behavior. Thirty-six licensed drivers completed a CARLA-based simulator study under one of three conditions: modular, where features required explicit activation within separate boundaries; integrated, where functions were bundled under one control logic and transitioned automatically; or hybrid, where longitudinal functions were combined while lateral control remained separate. All architectures offered adaptive cruise control, stop-and-go car-following, and lane centering assist. Longitudinal mode confusion did not differ significantly across architectures, but lateral mode confusion did. The integrated design showed the most behavioral conflicts, more abrupt lateral engagement, and the highest ADAS utilization. These findings highlight a tradeoff between interaction simplicity, boundary clarity, and driver awareness.