Generation of Real-time Synthetic Environment Using a Mobile Sensor Platform

Traditional methods used to generate synthetic environment databases for driving simulators often utilize automated software tools for the roadways combined with a lot of manual efforts for feature placement and fine-tuning. Automated tools, although they are much more efficient than manual techniques, have several shortcomings. They produce roadways that have a very regular and repeating appearance, in the sense that they contain no faults such as pavement cracks, potholes, asphalt runs, or any other such features that are so common in real roadways. Of course such "faults" can be added manually, but doing so in large databases is extremely time-consuming. Roadways generated by automated tools are often geometrically perfect, in the sense that super-elevation and crowning is applied consistently and smoothly, again with no faults typically encountered in real roadways. Finally, automated tools are often difficult to use when building geo-specific databases because the actual three-dimensional road geometry is not readily available. Our approach to addressing these concerns is to use advanced multi-modal sensor technology mounted on a vehicle that can survey a roadway while driving through it. The ultimate goal of the project is to use various sensors to obtain multiple-level-of-detail information about the road and the surrounding area, which can then be used to automatically produce synthetic environment databases for use in a driving simulator. Sensors include Global Positioning System (GPS) sensors for centimeter-level road geometry capture, scanning lasers for millimeter-level longitudinal and lateral road profile data, and various cameras combined with image processing techniques for automatic texture generation and feature identification. This paper describes the overall goals of the project, along with some preliminary results of using GPS receivers for survey of road geometry.