Abstract
M.Sc. (Mathematical Statistics)
This dissertation deals with the computation of nonlinear control strategies for an autonomous vehicle.
The vehicle consists of two wheels attached to an axle. It is assumed that both wheels roll without slipping leading to nonholonomic constraints. A third order nonlinear kinematic model of the vehicle is derived from these constraints. It is further assumed that the vehicle has builtin feedback controllers independently regulating the rotational velocities of the wheels (using electric motors as actuators). Thus, the vehicle is maneuvered by applying a separate rotational velocity reference command to the feedback controller of each wheel. The closed loop system dynamics from the reference command to the wheel rotational velocity is approximately modelled by a first order system. This leads to a fifth order nonlinear state-space model for the
vehicle. The above-mentioned reference commands constitute the control input variables of the vehicle model and are subject to amplitude constraints.
Firstly, a methodology is developed for computing reference command strategies to drive the autonomous vehicle from a specified initial state to a desired final state in a given time and such that a circular obstacle is avoided. The vehicle performs the required maneuver whilst satisfying all the specifications and constraints.
Secondly, feedback reference command laws are developed such that a specified point just ahead of the vehicle asymptotically tracks a given reference trajectory in the horizontal plane.
The feedback control law steers the vehicle onto the reference path from any initial position and keeps it moving on the path.
Thirdly, the stochastic system performance is evaluated when the above-mentioned tracking control strategy is applied and the initial state of the vehicle is a random vector.