Virtual Reality for Autonomous Vehicles: A Review of Safety, Training, and Human–Machine Interaction

Abstract

Virtual Reality (VR) enables immersive, repeatable, and risk-free simulation of haz-ardous driving scenarios. VR is increasingly applied in autonomous vehicle (AV) re-search for safety evaluation, takeover training, and human–machine interaction (HMI) prototyping. This review synthesizes empirical work to assess the contributions of VR to takeover performance, training efficacy, ecological validity, and user ac-ceptance. Evidence suggests that VR interventions can reduce takeover reaction times compared to conventional instruction, support iterative HMI design, and enhance user familiarity and calibrated trust in AV systems. Persistent limitations include simulator sickness, sample homogeneity, fidelity gaps between VR and on-road performance, and inconsistent reporting of scenario design. We recommend standardized reporting (PRISMA flow plus scenario metadata), longitudinal transfer studies, cross-cultural samples, and hybrid VR–AR validation methods to strengthen transferability and reg-ulatory acceptance.