Modeling virtual environments filled with autonomous traffic

This thesis proposes a new methodology for modeling road networks at the geometric level, with the purpose to reproduce real-world road networks with an appropriate level of accuracy as well as a low level of workload. After evaluating existing data sources and road network extraction approaches, coarse scale aerial images are selected as data sources of road networks on the 2D x-y plane. An image processing model is proposed to perform road categorization, road segmentation, and road segment centerline extraction on aerial images. The third component: elevation, which cannot be extracted from aerial images, is obtained from another data source: GIS, a MASS public geospatial database.;At the topological level, this thesis introduces topological structures to store all information about the virtual environment. Two inter-connected topological structures are proposed: 1) global topological structure and 2) local topological structure. Each component of the road network such as road segments, intersections, bridges, etc. is represented as a node in the global topological structure. Connectivity and traffic information is stored at the global level. For each global topological node, a local topological structure is associated. Each component of the global node is represented as a node in the local topological structure. Information about the local driving environment such as signs, obstacles, etc. is stored at the local level topological structure. The idea of having topological structures at different levels enables autonomous objects to plan and follow its global route in the road network with ability to react to its instantaneous local driving conditions.;The issue of implementing autonomous traffic is then addressed. A combination of a perception model, a communication model and a decision-making model is implemented to achieve the goal of modeling autonomous objects that behave intelligently. All models are attached to autonomous objects that represent people and vehicles in a virtual environment. This enables such an autonomous object to be an independent entity that is self-motivated and self-controlled. The perception model enables autonomous agents to perceive their instant driving situation. Autonomous agents communicate with other autonomous objects via their communication model and the decision-making model relies on the perception model and the communication model to investigate possible outcomes before making decisions. The communication model defines senders, receivers, contents, and channels (media through which content is transferred) in real-time to gather desired information from specified objects. The decision-making model is divided into two levels, the global level and the local level. These work, respectively, with environmental information perceived by the perception model and traffic information received by the communications model. A group of logic rules are formulated as decision trees to model the process of making decisions on the basis of real-time activities.;A database is used to store both geometric and topological information about the virtual environment. All information is then loaded into in-memory arrays for real-time access and update. Triggers and Colliders, which are attached to all simulated objects, are adopted in this project to update dynamic arrays in all events. Also, triggers and colliders play key roles in designing repeatable driving scenarios. (Abstract shortened by UMI.).