PACE-ADS is a human-centered framework for automated vehicles. It combines three AI agents: one interprets the road environment, one infers passenger states from signals such as facial expressions and interprets verbal commands, and one combines those inputs into high-level driving decisions.

In closed-loop CARLA simulations, the system responded selectively when passenger states or instructions changed, or when the vehicle became immobilized. Under prescribed passenger-state trajectories, it adjusted driving behavior along motion-related dimensions associated with comfort and supported recovery from immobilization through automated reasoning and guidance from the occupant.