Robotic manipulators are used in many different scenarios these days. If one of these manipulators is moved from one location to another it may require a total reprogramming of that manipulator because of the new environment the robot is working in. This is because the path planning and trajectory planning scheme which works in one environment might not be suitable in another. This text takes a look at how an intuitive and easy to use motion planning framework for finding paths in different static environments or scenarios can be made. The use of the Open Motion Planning Library has been used for the path planning and second-order cone programming solved by SeDuMi in Matlab has been used for finding the time-optimal trajectory.

Two configurations for a unicycle, a class of nonholonomic systems, is looked into. The first configuration is the case where the generalized coordinates consists of four parameters, while the other configurations consists of five. For both systems the equations of motion are calculated. Then a method for parameterizing a desired path for the systems using a synchronization function among all degrees of freedom is shown. A set of equations to find feasible trajectories keeping the desired path virtually constrained is calculated. It is then shown for the four degree of freedom system how to compute the transverse linearization, which can be used for orbital stabilization of a desired motion consisting of circular orbits or straight paths. A way to compute a periodic controller for a linear time varying system is derived, and a set of controllers are tried out on the four degree of freedom unicycle with negative result.

In Augmented Reality, computer generated information can assist the userwith 'invisible' information about real world objects. In this thesis, thepossibilities for Augmented Reality on mobile devices will be investigated,and Object Tracking is identied as a vital functionality. In the existing so-lutions, position- and orientation-based tracking is most common, however,this project will approach vision-based tracking, which is a wide researcharea in the eld of Computer Vision.The chosen object tracker, is the prize-awarded algorithm TLD, which together with the recently released FastCVlibrary, makes up the functionality in the nal presented application. Theapplication is tested with regard to memory and processing time, and it is observed from the experiments, that the Detector module in TLD domi-nates the processing time. Nonetheless, the nal Object Tracker performsas intended and decent frame rates is obtained by reducing the size of thecamera image frame.For further work, the processing time could be reduced even more, possiblyby making a longer interval between each run with the detector module.Also the possibility to save object models and use them later on should beexplored. This would be useful in an Augmented Reality application, wherethe objects usually is predetermined.