A Hybrid Driving Simulator with Dynamics-Driven Motion and Data-Driven Motion

Abstract

In driving simulator technology, which reproduces the sense of motion in a virtual reality environment, various research activities are attempting to enhance realism. Recently studies have focused on a data-driven method of generating motion; this method records the motions of real objects and regenerates them in a simulation system. By providing a simulator with real motion data, this method can easily guarantee realism without complicated dynamics models and equation solvers. This paper proposes a hybrid method of generating controllable motion in a driving simulator through the application of a data-driven motion generation method and a dynamics-based motion generation method. The driving data acquired from a real vehicle are parsed into motion blocks with a dynamic model of the vehicle; the motion blocks are then stored in the database. In real-time simulation, a natural motion stream is generated by a process of searching for and synthesizing optimal motion blocks from the database and is synthesized with dynamics-based motion again. A more realistic sense of motion can be generated by parameterizing the current simulation state and user requirements.