| This dissertation studies issues in the implementation of the control system and path-planning algorithms in autonomous ground vehicles (AGV), which are designed for and further developed from both DARPA Grand Challenges in 2004 and 2005.;A hierarchical generic control architecture of the AGV control system is presented. In this architecture, the functionalities required for an AGV control system are decomposed into small and well defined tasks which are then carried out by different controllers. As an example, ION, the Intelligent Off-road Navigator, the autonomous vehicle participated in DARPA Grand Challenge 2005, is then introduced. Its control system structure and the hierarchical task decomposition are studied and designed. The discrete-event state machine in ION's navigation module and its physical layout of the vehicle hardware is also discussed.;In order to safely, smoothly and efficiently guide the vehicle to the goals, the navigation module in the control system is a very important part. The path-planning algorithms applied in ION's navigation module are then elaborated. We develop a real-time path-planning algorithm utilizing fuzzy logic. Two different fuzzy controller, fuzzy steering controller and fuzzy speed controller, are designed and discussed to ensure the planned path is goal-reaching and obstacle-free. In addition, because of the vehicle's physical constraints, we study the smooth path planning problem for the autonomous ground vehicles and the quartic spline interpolation method is proposed. In order to make sure that the distance between the line segments of the given waypoints and the smooth path is small so that the vehicle stays right on the path, we formulate and study the minimum offset smooth interpolation problem. With the given way points and maximum turning curvature, the minimum offset can be calculated by using quartic splines. In the end we consider the maneuver planning problem. The vector space and state flow of the nonholonomic systems are defined. Three different approaches, Lie algebra approximation, dynamic programming, and nonlinear control method, are presented for different applications. |
