No description available.

No description available.

No description available.

No description available.

This thesis presents a framework for formal system development. The framework is called `RD' which is short for `Reasoning about Designs'. RD integrates proof, development and diagnostic modes of reasoning. Many commonly studied formalisms are shown to be consistent with this framework. A large example based on an industrial problem is given to demonstrate RD. The integration of system design and management is achieved by unifying formal software engineering methods and model-based reasoning. RD formally specifies a complete toolkit for performing system development and then re-using the development as the system description for diagnostic reasoning. RD does not restrict the contributing system analysis methods, rather it maps out and defines the entities and relations common to both. The framework is, in principle, extensible to support other forms of reasoning. The ground technical mechanism of the framework is a novel view of formal system development based on a general implementation relation. Implementation relations are widely studied in formal methods in software engineering where they are often referred to as `refinement'. RD allows refinement relations to be defined in a way that makes expected behaviours and faults of system implementations explicit. Furthermore, a case is made that all well known forms of refinement implicitly support diagnostic reasoning as they can be restated within the framework. RD is an integrated and completely rigorous approach to the core system building tasks of design and management. Despite the large amount of technical detail, the following discussion can be seen as raising many issues that relate to engineering in general. In particular, a formal engineering process should have benefits beyond just the delivery of systems that satisfy their specifications.

In many image-processing applications, one object of interest must be segmented. The techniques used for segmentation vary depending on the particular situation and the specifications of the problem at hand. In methods that rely on a learning process, the lack of a sufficient number of training samples is usually an obstacle, especially when the samples need to be manually prepared by an expert. The performance of some other methods may suffer from frequent user interactions to determine the critical segmentation parameters. Also, none of the existing approaches use online (permanent) feedback, from the user, in order to evaluate the generated results. Considering the above factors, a new multi-stage image segmentation system, based on Reinforcement Learning (RL) is introduced as the main contribution of this research. In this system, the RL agent takes specific actions, such as changing the tasks parameters, to modify the quality of the segmented image. The approach starts with a limited number of training samples and improves its performance in the course of time. In this system, the expert knowledge is continuously incorporated to increase the segmentation capabilities of the method. Learning occurs based on interactions with an offline simulation environment, and later online through interactions with the user. The offline mode is performed using a limited number of manually segmented samples, to provide the segmentation agent with basic information about the application domain. After this mode, the agent can choose the appropriate parameter values for different processing tasks, based on its accumulated knowledge. The online mode, consequently, guarantees that the system is continuously training and can increase its accuracy, the more the user works with it. During this mode, the agent captures the user preferences and learns how it must change the segmentation parameters, so that the best result is achieved. By using these two learning modes, the RL agent allows us to optimally recognize the decisive parameters for the entire segmentation process.

Vehicle tracking in environments containing occlusion and clutter is an active research area. The problem of tracking vehicles through such environments presents a variety of challenges. These challenges include vehicle track initialization, tracking an unknown number of targets and the variations in real-world lighting, scene conditions and camera vantage. Scene clutter and target occlusion present additional challenges. A stochastic framework is proposed which allows for vehicles tracks to be identified from a sequence of images. The work focuses on the identification of vehicle tracks present in transportation scenes, namely, vehicle movements at intersections. The framework combines background subtraction and motion history based approaches to deal with the segmentation problem. The tracking problem is solved using a Monte Carlo Markov Chain Data Association (MCMCDA) method. The method includes a novel concept of including the notion of discrete, independent regions in the MCMC scoring function. Results are presented which show that the framework is capable of tracking vehicles in scenes containing multiple vehicles that occlude one another, and that are occluded by foreground scene objects.

This research explores the development of vibration-based haptic feedback for a mouse-like computer input device. The haptic feedback is intended to be used in 3D virtual environments to provide users of the environment with information that is difficult to convey visually, such as collisions between objects. Previous research into vibrotactile haptic feedback can generally be split into two broad categories: single tactor handheld devices; and multiple tactor devices that are attached to the body. This research details the development of a vibrotactile feedback device that merges the two categories, creating a handheld device with multiple tactors. Building on previous research, a prototype device was developed. The device consisted of a semi-sphere with a radius of 34 mm, mounted on a PVC disk with a radius of 34 mm and a height of 18 mm. Four tactors were placed equidistantly about the equator of the PVC disk. Unfortunately, vibrations from a single tactor caused the entire device to shake due to the rigid plastic housing for the tactors. This made it difficult to accurately detect which tactor was vibrating. A second prototype was therefore developed with tactors attached to elastic bands. When a tactor vibrates, the elastic bands dampen the vibration, reducing the vibration in the rest of the device. The goal of the second prototype was to increase the accuracy in localizing the vibrating tactor. An experiment was performed to compare the two devices. The study participants grasped one of the device prototypes as they would hold a computer mouse. During each trial, a random tactor would vibrate. By pushing a key on the keyboard, the participants indicated when they detected vibration. They then pushed another key to indicate which tactor had been vibrating. The procedure was then repeated for the other device. Detection of the vibration was faster (p < 0.01) and more accurate (p < 0.001) with the soft shell design than with the hard shell design. In a post-experiment questionnaire, participants preferred the soft shell design to the hard shell design. Based on the results of the experiment, a mould was created for building future prototypes. The mould allows for the rapid creation of devices from silicone. Silicone was chosen as a material because it can easily be moulded and is available in different levels of hardness. The hardness of the silicone can be used to control the amount of damping of the vibrations. To increase the vibration damping, a softer silicone can be used. Several recommendations for future prototypes and experiments are made.

In many image-processing applications, one object of interest must be segmented. The techniques used for segmentation vary depending on the particular situation and the specifications of the problem at hand. In methods that rely on a learning process, the lack of a sufficient number of training samples is usually an obstacle, especially when the samples need to be manually prepared by an expert. The performance of some other methods may suffer from frequent user interactions to determine the critical segmentation parameters. Also, none of the existing approaches use online (permanent) feedback, from the user, in order to evaluate the generated results. Considering the above factors, a new multi-stage image segmentation system, based on Reinforcement Learning (RL) is introduced as the main contribution of this research. In this system, the RL agent takes specific actions, such as changing the tasks parameters, to modify the quality of the segmented image. The approach starts with a limited number of training samples and improves its performance in the course of time. In this system, the expert knowledge is continuously incorporated to increase the segmentation capabilities of the method. Learning occurs based on interactions with an offline simulation environment, and later online through interactions with the user. The offline mode is performed using a limited number of manually segmented samples, to provide the segmentation agent with basic information about the application domain. After this mode, the agent can choose the appropriate parameter values for different processing tasks, based on its accumulated knowledge. The online mode, consequently, guarantees that the system is continuously training and can increase its accuracy, the more the user works with it. During this mode, the agent captures the user preferences and learns how it must change the segmentation parameters, so that the best result is achieved. By using these two learning modes, the RL agent allows us to optimally recognize the decisive parameters for the entire segmentation process.

Vehicle tracking in environments containing occlusion and clutter is an active research area. The problem of tracking vehicles through such environments presents a variety of challenges. These challenges include vehicle track initialization, tracking an unknown number of targets and the variations in real-world lighting, scene conditions and camera vantage. Scene clutter and target occlusion present additional challenges. A stochastic framework is proposed which allows for vehicles tracks to be identified from a sequence of images. The work focuses on the identification of vehicle tracks present in transportation scenes, namely, vehicle movements at intersections. The framework combines background subtraction and motion history based approaches to deal with the segmentation problem. The tracking problem is solved using a Monte Carlo Markov Chain Data Association (MCMCDA) method. The method includes a novel concept of including the notion of discrete, independent regions in the MCMC scoring function. Results are presented which show that the framework is capable of tracking vehicles in scenes containing multiple vehicles that occlude one another, and that are occluded by foreground scene objects.