| A major goal in robotics is to develop machines that perform useful tasks with minimal supervision. Instead of requiring each small detail to be specified, one would like to describe the task at a high level, and have the system autonomously execute it correctly.;The work presented in this thesis addresses some of the challenges of creating autonomous systems that perform complex tasks while providing guarantees for correct behavior. Here, tasks defined using the formalism of temporal logic are automatically translated into controllers that drive a robot or group of robots and activate the different robot actions such that the task is completed, if it is feasible. The approach presented is correct by construction, it is complete in the sense that if under specific assumptions the task can be done, a controller will be created and it is sound, that is, if the task cannot be accomplished the algorithms return that it is infeasible and no controller is created.;The approach to generating such controllers comprises of three steps. First, the task itself is described; This description captures the required behavior as well as relevant information about the robot capabilities such as actions and sensors, and the environment in which the robot is operating. This step transforms a generally continuous problem into a discrete one by creating abstractions of the different task components. The next step is to find a discrete solution for the problem, a discrete plan, that if followed the robot is guaranteed to achieve the discrete behavior required of it. The final step is to return to the original continuous problem by transforming the discrete plan to a hybrid controller that switches between a set of atomic controllers or control primitives that provide continuous local guarantees for the behavior of the robot. |
