| This thesis addresses two core aspects of building virtual urban environments for real-time simulation: roadway modeling and behavior control of virtual vehicles.; Roadways are more than polygons used for rendering the scene. In order to support behavior control of agents efficiently, we must describe the road network according to requirements of behavior control. We describe a road network as a ribbon network. A ribbon defines the geometry and orientation of a 3D navigable surface. A ribbon provides spatial information for tracking and a frame of reference for local spatial relationships among the agents.; A parametric piecewise cubic spline curve is used to model the central axis of a ribbon. However, in general, the equal parameter intervals don't correspond to equal arc-length intervals. This thesis presents a method to compute an approximately arc-length parameterized cubic spline curve to model the central axis of a ribbon.; On a ribbon; we defined a curvilinear coordinate system. Mapping between 3D Cartesian and local ribbon coordinates is often a bottleneck in real-time simulations. A main problem in mapping computations is to compute, given a point in 3D space; the closest point on a spatial curve. We create an efficient, extremely robust two-stage method which takes complementary strengths of two optimization methods to solve this problem.; We introduce a component-wise method to control behaviors of autonomous agents driving on ribbon based road networks. Behavior control is divided into tracking, cruising, following, intersection, and lane changing behaviors. The component behaviors are then combined to generate values for control parameters. A data structure called a path is created for controlling these steering behaviors.; For intersection behavior, the topology of intersections and right-of-way rules and social conventions on intersections are embedded into the environment database. Intersection behavior controls driving agents to follow rules and conventions when they cross intersections.; Lane changing behavior controls a virtual vehicle to leave the original lane and enter an adjacent lane. We create another higher level data structure called a route which captures more global intentions and guides local choices about the immediate path. |
