22 Automated Static and Dynamic Obstacle Avoidance in Arbitrary 3D Polygonal Worlds J.M.P. van Waveren and L.J.M. Rothkrantz Delft University of Technology, Netherlands Defence Academy The Netherlands 1. Introduction In common applications, a robot will have no prior knowledge of its environment and must use sensory information to build a cognitive model for path and route finding. The process of building a good cognitive model from sensory information alone is complex and time consuming. Uncertainty of the input data can also cause the resulting paths and routes to be suboptimal or erroneous. In many of these applications, however, it is possible to augment a robot's sensory information with existing static information about the environment. Detailed and accurate digital blueprints of buildings are a fortuitous by-product of the prevalence of computer aided design in architectural design today. These data provide some information about the layout of the environment, but typically do not include more dynamic elements such as furniture, people, or other robots. Additional records may indicate what type of furniture is in a given room, but its placement may have changed over time. A robot navigating an office building will have to deal with the uncertainties of dynamic and unexpected objects in real-time, even if the layout of the building is known. However, the situation is vastly improved if the sensory data is used to identify object types, whose precise dimensions are already known and can be retrieved as needed. The use of existing information of the environment is important to make a robot resource- efficient, while achieving near optimal path and route finding. However, processing existing information in the form of geometric models with hundreds of thousands of polygons can be expensive if a robot system is not carefully crafted to deal with such large amounts of information. An automated system is presented for path and route finding through arbitrary 3D polygonal environments. The system can process a polygonal representation of an environment with hundreds of thousands of polygons within a few minutes on a small grid of today’s computers. The processed information allows a robot to efficiently find routes and paths through the environment in real-time. Additionally, when a robot has identified and recognized dynamic obstacles in the environment based on sensory information, the system is able to create near optimal paths in real-time around arbitrary configurations of dynamic obstacles. The system presented here has been battle hardened in several generations of computer games, such as the triple-A titles: QUAKE III Arena, DOOM III and Enemy Territory QUAKE Wars. These computer games provide virtual environments for a robot (or artificial player), which are by no means inferior to real-life environments in that sophisticated www.intechopen.com