Autonomous-vehicle (AV) technologies are rapidly advancing, but a great deal remains to be learned about their interaction and perception on public roads. Research in this area usually relies on AV trials using naturalistic driving which are expensive with various legal and ethical obstacles designed to keep the general public safe. The emerging concept of Wizard-of-Oz simulation is a promising solution to this problem wherein the driver of a standard vehicle is hidden from the passenger using a physical partition, providing the illusion of riding in an AV. Furthermore, head-mounted display (HMD) virtual reality (VR) has been proposed as a means of providing a Wizard-of-Oz protocol for on-road simulations of AVs. Such systems have potential to support a variety of study conditions at low cost, enabling simulation of a variety of vehicles, driving conditions, and circumstances. However, the feasibility of such systems has yet to be shown. This study makes use of a within-subjects factorial design for examining and evaluating a virtual reality autonomous vehicle (VRAV) system, with the aim of better understanding the differences between stationary and on-road simulations, both with and without HMD VR. More specifically, this study examines the effects on user experience of conditions including presence, arousal, simulator sickness and task workload. Participants indicated a realistic and immersive driving experience as part of subjective evaluation of the VRAV system, indicating the system is a promising tool for human-automation interaction and future AV technology developments.