| Abstract |
In this research, we explore the use of virtual reality (VR) to elevate trust towards self-driving cars (SDCs), as safety concerns remain a major obstacle in their widespread acceptance. By simulating two dangerous traffic scenarios in a virtual world, participants experience being in a self-driving car and observe SDCs' informed decisions without potential harm or accidents. To realistically simulate these scenarios, we utilized several machine learning techniques, namely object detection, reinforcement and imitation learning, to accurately identify obstacles and interpret traffic lights' colors, allowing the self-driving car to make informed decisions while navigating to avoid road accidents. To evaluate how our VR simulation would elevate user's trust, we conducted an experiment that is composed of two parts. The first part involved participants driving a regular car within the VR environment, while the second part involved participants experiencing being inside a self-driving car autonomously making its own decisions. The results showed that participants expressed trust rates in the self-driving car that are comparable to that of the regular one. Moreover, most participants did not feel the need to take control at any point while the SDC was avoiding the accident or stopping at the red light. We believe that our work is a step towards fostering trust and acceptance of SDCs, which advances their integration into smart cities. © 2023 IEEE. |
