Abstract: [Purpose/Significance] Technological breakthroughs in artificial intelligence have accelerated the process of autonomous driving system, ushering in another round of eager attention to the autonomous driving industry and related industries. Augmented reality head-up display（AR-HUD）is currently one of the most promising human-AI interfaces in the field of autonomous driving.Exploring which type of AR-HUD visual explanation can improve user acceptance of autonomous driving system under different environmental visibility conditions will help improve the transparency of decision-making based on autonomous driving system and promote the development of explainable artificial intelligence（Explainable AI）. [Design/Methodology] This controlled experiment adopted a within-subjects design of （3 visual explanation type: icon/text/none）× 2（environmental visibility: high/low）, with 61 participants in 6 different simulated scenarios of driving interacted with AR-HUD.The human-AI interaction experience with scales were measured, including the user's perception, attitude, and willingness to the autonomous driving system. [Findings/Conclusion] The type of visual explanation of AR-HUD significantly affects user acceptance of autonomous driving system, and icon explanation's effect is significantly better than text explanation and no explanation; however, there is no significant moderating effect of environmental visibility on user acceptance. [Originality/Value] By combining simulation software and physical models, this study creates realistic driving scenes and accurately measures the human-AI interaction experience through controlled experiments, which provides an empirical basis for opening the 'blackbox' of autonomous driving by AR-HUD design, and proposes a methodological supporting for measuring the human-AI interaction experience.

Keywords: Human-AI interaction, Autonomous driving, Explainable, Acceptance, Augmented Reality Head-Up Display