| This dissertation describes an architecture for a Path Planner system that guides an autonomous ground vehicle through an urban environment while obeying a basic set of traffic rules. As urban driving requires the handling an open-ended set of scenarios, the planner is designed to address the scenarios specified by DARPA's Urban Challenge (UC); however, the planner permits easy addition of new capabilities to address more scenarios in the future.;Each capability required for urban driving is achieved separately by individual planners called Basic-Planners (BPs). A sequence of these BPs are used by the rest of the Path Planner to achieve the complete mission. The specific type of the BP is kept oblivious from rest of the system to provide easy plug-ability of these BPs. The dissertation introduces the BPs required to match the requirements of UC, and the architecture is designed to provide this plug-ability.;It also introduces a novel approach for path planning in open-areas such as a parking-lot. This planner is created by merging two classical approaches, namely, grid-based planning and Dubin's based continuous space planning.;The system is then evaluated by specifying its performance at the UC, where system is used to guide CajunBot-II. It then provides a comparison of the present architecture with that of the three of the finalist of UC, namely, Boss, Junior and Skynet. |
