This thesis presents a method for performing tracking and estimation of head position and orientation by means of template based particle filtering. The implementation is designed to withstand high levels of occlusion and noise, and allow for system dynamics to be accounted for. To accelerate the computation, GPGPU techniques are used to enable the GPU to function a co-processor, resulting in real-time performance. A method is devised for dynamic creation of feature points used in the particle filter. Furthermore, the graphics pipeline is used to overlay and visualize the tracking, as well as play a key role in the dynamic template functionality. Finally, a benchmarking system is suggested and developed for carrying out controlled evaluation of tracking methods in general.

The autonomous bike was conceived by Jens G. Balchen back in the eighties, and later picked up by Amund Skavhaug. The idea of a two-wheeled, self-powered, yet riderless bike has since been pursued intermittently over the years. Audun S{o}lvberg was the last person working on the project in 2007, making large headway towards the final goal of the bike: An outdoor ride, preformed by a riderless bicycle.The same goal is shared by this thesis; to make the bike work as intended and get the bicycle running outside. Additionally, as it would be naive to assume that this work would leave the bike in perfect condition, the thesis focuses on documenting issues that are left unfixed upon completion. Unfortunately, a fully functional, outdoor demonstration was not achieved, but half of the battle was won: Two of the three motors residing on the bike can be fully controlled by an external device.At the start of the work, the instrumentation system was mostly completed. A computer, running a QNX Neutrino Operating System, interfacing with potmeters, motors, Inertial Measurement Units and a GPS where available, mounted on the bike. Drivers communicating with the hardware was already written, and a Simulink model, meant to control the bike had been developed. However, the motherboard was in need of replacement, and the Simulink model was not finalized or even tested. The system was also lacking wireless networking capabilities.Yet, at the onset of the thesis, it was believed that the project was very close to reaching its ultimate goal.As the work progressed, several issues became apparent, emerging along with problems being solved. The motherboard was replaced with a motherboard{}, implicitly requiring a new hard drive to be installed. The OS was upgraded from version 6.3.0 to 6.4.1, to bring it up to date and as a requirement for some of the motherboard hardware. All of the device drivers where modified to work without their counterpart hardware connected, easing development of the Simulink model. The model was shown to be unreliable, but the hardware interface subsystem was completed and tested to allow for easy integration in separate projects. To demonstrate the capabilities of subsystem, a Bike Demo model was created, allowing the steer and pendulum to be controlled by the used of the bike accelerometer, using orientation data as reference. The intent was for the demo to serve as a control system for an outdoor ride. Ultimately the motor controller card was found to be incompatible with the propulsion motor, unable to deliver sufficient current. At the time of discovery, too little time reminded for the problem to be rectified. A demonstration of control system however, was successfully concluded.An attempt to bring wireless networking to the bike failed. Arguably too much time was spent making USB Wifi devices working with the QNX OS, when easier alternatives could perhaps have solved the problem earlier.The thesis has a large focus on the work ahead, as the system is complex and unreliable by its very nature. All known issues are detailed and summarized, and the various chapters describing hardware and software have been outlined in an attempt to serve as a go-to reference for students taking on the autonomous bike project in the future.

This thesis describes the implementation of a Time Management Unit (TMU) in hardware as specified by Gregertsen and Skavhaug, the specification and implementation of several improvements to the proposed specification, and the creation of a software framework to enable use of the module in a convenient way. A set of thorough automatic functional tests are also described and provided. The performance of the module is assessed and discussed. A user description similar to the AVR32 UC3 datasheets is also created.The TMU has been implemented as a hardware module on the peripheral bus (APB) on the AVR32 UC3 microcontroller, which makes it easy to develop and test stand-alone, and simple to integrate into future UC3 microcontrollers. Also, as the APB interface of the AMBA standard is an open standard used by several System-on-a-chip (SoC) designs, the module can be implemented on other microcontrollers with very low effort.The final product makes it possible to measure and control the execution time of tasks with high precision and low overhead. It supports atomic swapping of registers in a manner closely related to a context switch.Gregertsen and Skavhaugs’s research in implementing support for the Ada language and run-time environment on the UC3 microcontroller will benefit directly from this project, as the system relies on the hardware support provided by the TMU. Also, as the project can be used in proving that hardware support of execution time monitoring may allow for new ways of ensuring schedulability in real-time systems, it can possibly be a part of a new direction in real-time research.

The idea of an autonomous bicycle originates from Jens G. Balchen who wanted to make an unmanned autonomous bicycle. The idea was picked up by Amund Skavhaug who extended the idea with the concept of using an inverted pendulum to simulate a leaning rider. The previous attempts to develop a bicycle capable of performing an autonomous ride has so far all ended in failure. The main reason for the Department of Engineering Cybernetics is to develop such a bicycle is for use in recruitment and motivation of students. The main goal of this thesis is to develop a bicycle that after the implementation of a suitable control is capable of performing an autonomous ride.The goal of this thesis is to create a controller making the bicycle capable of performing the first self balanced ride. The focus is not on implementing the most advanced controller but creating a system actually capable of performing this first ride. An equally important focus is that the framework delivered at the end of this thesis is capable of handling the further development towards a fully autonomous bicycle.The author has during this thesis performed the additions needed in order to be able to implement a self balancing controller. The parameters of the real bicycle were measured and used to create a simulation environment of the bicycle in Simulink. Several controllers were simulated in Simulink, before a controller were implemented on the physical bicycle.The physical bicycle delivered as a part of this thesis consists of a fully functional framework both capable and ready for the further development. A self balancing controller is implemented on the bicycle and the bicycle has performed it's first self balance ride.