On Autonomous Articulated Vehicles

Abstract: The objective of this thesis is to address the problems of modeling, path planning and path following for an articulated vehicle in a realistic environment and in the presence of multiple obstacles.In greater detail, the problem of the kinematic modeling of an articulated vehicle is revisited through the proposal of a proper model in which the dimensions and properties of the vehicle can be fully described, rather than considering it as a unit point. Based on this approach, nonlinear and linear error kinematics models are proposed that are also able to account for the effect of the slip angles, a factor that can cause dramatic deterioration in the overall performance of the vehicle.Subsequently, two different concepts for addressing the problem of path following for articulated vehicles are proposed. The first concept is based on a switching model predictive control architecture, which relies on multiple switching linear error dynamics models of the articulated vehicle to account for the effect of varying the slip angles and cruising speed as well as the mechanical and physical constraints of the vehicle.The second proposed control concept is a novel nonlinear sliding mode controller that introduces continuous sliding surfaces to reduce chattering effects while tracking a reference trajectory.The sliding mode controller is utilized based on the extracted nonlinear error coordinates of the articulated vehicle. The feasibility of this approach has been demonstrated through multiple experimental tests on a small scale using a fully realistic articulated vehicle.Finally, in the path planning part of the thesis, artificial potential field and bug algorithms are addressed. More specifically, the potential field path planning algorithm is modified by considering the nonlinear kinematic model of the articulated vehicle and correspondingly adapting the repulsive and attractive coefficients.In the case of the well-known bug algorithm, a suitable navigation method for an articulated vehicle for local path planning based on a minimum set of sensors and with decreased complexity for online implementation is also proposed.Furthermore, the performance of the modified potential field method has been experimentally evaluated in multiple path planning scenarios using the previously mentioned small-scale realistic articulated vehicle.