HTC Vive Pro Time Performance Benchmark for Scientific Research

Abstract

Widespread availability of consumer-level virtual reality (VR) devices creates a venue for their massive use in psychology and neuroscience research. The application of VR to scientific research however poses significant constraints on system performance and stability. In particular, studies with multimodal measurement of human behavior and physiology require precise hardwaresoftware synchronization with precise event labeling (within 10 milliseconds). Previous works investigating suitability of VR systems for research have mainly focused on benchmarking performance in spatial tracking. Therefore, it remains unclear if timing parameters such as latency or jitter in VR motion capture and VR audiovisual stimulation allow for carrying out science under strong time constraints. Here we present the first quantitative test of time performance in VR input and VR feedback of the current state-of-the-art HTC Vive Pro system. Using both low-level Python-based API and a high-level game engine (Unity), our multilevel testing procedure allows us to isolate software influence on observed results. We report that, in both test conditions, latencies are non-negligible considering fine synchronization with multimodal measurements; however, jitters are stable and low, which allows to counter-balance the effect of latency by using constant offsets to re-synchronize multimodal data. Finally, we plan to share our testing hardware setup as an open-source and low-cost benchmark toolkit, allowing objective testing to be easily reproduced by the community in an open collaborative framework.