The purpose of the research presented in this dissertation is to improve virtual reality (VR) training systems by enhancing their understanding of users. While the field of intelligent tutoring systems (ITS) has seen value in this approach, much research into making use of biometrics to improve user understanding and subsequently training, relies on specialized hardware. Through the presented research, I show that with machine learning (ML), the VR system itself can serve as that specialized hardware for VR training systems. I begin by discussing my explorations into using an ecologically valid, specialized training simulation as a testbed to predict knowledge acquisition by users unfamiliar with the task being trained. Then I look at predicting the cognitive and psychomotor outcomes retained after a one week period. Next I describe our work towards using ML models to predict the transfer of skills from a non-specialized VR assembly training environment to the real-world, based on recorded tracking data. I continue by examining the identifiability of participants in the specialized training task, allowing us to better understand the associated privacy concerns and how the representation of the data can affect identifiability. By using the same tasks separated temporally by a week, we expand our understanding of the diminishing identifiability of user's movements. Finally, I make use of the assembly training environment to explore the feasibility of across-task identifiability, by making use of two different tasks with the same context.

Augmented Reality's (AR) scope and capabilities have grown considerably in the last few years. AR applications can be run across devices such as phones, wearables, and head-mounted displays (HMDs). The increasing research and commercial efforts in HMDs capabilities allow end users to map a 3D environment and interact with virtual objects that can respond to the physical aspects of the scene. Within this context, AR is an ideal format for in-situ training scenarios. However, building such AR scenarios requires proficiency in game engine development environments and programming expertise. These difficulties can make it challenging for domain experts to create training content in AR. To combat this problem, this thesis presents strategies and guidelines for building authoring tools to generate scenario-based training experiences in AR. The authoring tools were built leveraging concepts from the 3D user interfaces and interaction techniques literature. We found from early research in the field and our experimentation that scenario and object behavior authoring are substantial aspects needed to create a training experience by an author. This work also presents a technique to author object component behaviors with high usability scores, followed by an analysis of the different aspects of authoring object component behaviors across AR, VR, and Desktop. User studies were run to evaluate authoring strategies, and the results provide insights into future directions for building AR/VR immersive authoring tools. Finally, we discuss how this knowledge can influence the development, guidelines, and strategies in the direction of a more compelling set of tools to author augmented reality SBT experiences.

Emerging research is expanding the idea of using 360-degree spherical panoramas of real-world environments for use in "360 VR" experiences beyond video and image viewing. However, most of these experiences are strictly guided, with few opportunities for interaction or exploration. There is a desire to develop experiences with cohesive virtual environments created with 360 VR that allow for choice in navigation, versus scripted experiences with limited interaction. Unlike standard VR with the freedom of synthetic graphics, there are challenges in designing appropriate user interfaces (UIs) for 360 VR navigation within the limitations of fixed assets. To tackle this gap, we designed RealNodes, a software system that presents an interactive and explorable 360 VR environment. We also developed four visual guidance UIs for 360 VR navigation. The results of a pilot study showed that choice of UI had a significant effect on task completion times, showing one of our methods, Arrow, was best. Arrow also exhibited positive but non-significant trends in average measures with preference, user engagement, and simulator-sickness. RealNodes, the UI designs, and the pilot study results contribute preliminary information that inspire future investigation of how to design effective explorable scenarios in 360 VR and visual guidance metaphors for navigation in applications using 360 VR environments.

The COVID-19 virus disrupted all aspects of our daily lives, and though the world is finally returning to normalcy, the pandemic has shown us how ill-prepared we are to support social interactions when expected to remain socially distant. Family members missed major life events of their loved ones; face-to-face interactions were replaced with video chat; and the technologies used to facilitate interim social interactions caused an increase in depression, stress, and burn-out. It is clear that we need better solutions to address these issues, and one avenue showing promise is that of Interpersonal Telepresence. Interpersonal Telepresence is an interaction paradigm in which two people can share mobile experiences and feel as if they are together, even though geographically distributed. In this dissertation, we posit that this paradigm has significant value in one-to-one, asymmetrical contexts, where one user can live-stream their experiences to another who remains at home. We discuss a review of the recent Interpersonal Telepresence literature, highlighting research trends and opportunities that require further examination. Specifically, we show how current telepresence prototypes do not meet the social needs of the streamer, who often feels socially awkward when using obtrusive devices. To combat this negative finding, we present a qualitative co-design study in which end users worked together to design their ideal telepresence systems, overcoming value tensions that naturally arise between Viewer and Streamer. Expectedly, virtual reality techniques are desired to provide immersive views of the remote location; however, our participants noted that the devices to facilitate this interaction need to be hidden from the public eye. This suggests that 360$^\circ$ cameras should be used, but the lenses need to be embedded in wearable systems, which might affect the viewing experience. We thus present two quantitative studies in which we examine the effects of camera placement and height on the viewing experience, in an effort to understand how we can better design telepresence systems. We found that camera height is not a significant factor, meaning wearable cameras do not need to be positioned at the natural eye-level of the viewer; the streamer is able to place them according to their own needs. Lastly, we present a qualitative study in which we deploy a custom interpersonal telepresence prototype on the co-design findings. Our participants preferred our prototype instead of simple video chat, even though it caused a somewhat increased sense of self-consciousness. Our participants indicated that they have their own preferences, even with simple design decisions such as style of hat, and we as a community need to consider ways to allow customization within our devices. Overall, our work contributes new knowledge to the telepresence field and helps system designers focus on the features that truly matter to users, in an effort to let people have richer experiences and virtually bridge the distance to their loved ones.

As time goes on, the demand for higher resolution and more visually rich images only increases. Unfortunately, creating these more realistic computer graphics is pushing our computational resources to their limits. In realistic rendering, one of the common ways 3D objects are represented is as volumetric elements called voxels. Traditionally, voxel data structures are known for their high memory requirements. One of the standard ways these requirements are minimized is by storing the voxels in a sparse voxel octree (SVO). Very recently, a method called High Resolution Sparse Voxel DAGs was presented that can store binary voxel data orders of magnitudes more efficiently than SVOs. This memory efficiency is achieved by converting the tree into a directed acyclic graph (DAG). The method was also shown to have competitive rendering performance to recent GPU ray tracers. Unfortunately, it does not support storing collections of rendering attributes, commonly called materials. These represent a given object's reflectance properties, and are necessary for calculating its perceived color. We present a method for connecting material information to High Resolution Sparse Voxel DAGs for mid-level scenes, with multiple meshes, and several different materials. This is achieved using an extended Sparse Voxel DAG, called a Moxel DAG, and an external data structure for holding the material information, we call a Moxel Table. Our method is much more memory efficient than traditional SVOs, and only increases in efficiency in comparison when at higher resolutions. Because it stores the equivalent information as SVOs, it achieves the exact same visual quality at the same resolutions.

We live in an age when consumers can now shop and browse the web using hand-held devices. This means that competitive companies need to have a website to represent their brand and to conduct business. E-commerce sites need to pay special attention to the usability of their sites, since it has such an impact on how potential costumers view their brand. Jakob Nielsen defines usability as a "quality attribute that assesses how easy user interfaces are to use"; he separates usability into five quality components: learnability, efficiency, memorability, errors and satisfaction. The current standard for testing usability involves having a number of users physically use a site in order to determine where they have trouble. This kind of usability testing can be time consuming and costly. In order to mitigate some of these costs, many tools are being developed to help automate the process. However, many automated tools evaluate only one of the five components, or simply look for errors. In an attempt to increase the reliability and scope of such testing, this paper investigates the effectiveness of automated usability evaluators and proposes methods for future researchers to test them. Specifically, this paper details an experiment performed to test the some freely available usability evaluators against more traditional usability evaluations. The experiment attempts to determine whether automatic usability evaluations might be used as a cheaper alternative to more traditional usability evaluations.

No description available.

No description available.

This research assessed the effects of inconsistencies in graphic direct manipulation interfaces. Objective and subjective measurement techniques were employed to determine how inconsistencies affected performance in an Apple Macintosh-based computer application called “The Personal Organizer.” Three groups of 11 participants, all familiar with the Macintosh computer, were given a set of similar tasks on different versions of the application in a pretest (control version), treatment (control or one of two inconsistent versions), post-test (control version) experimental design. Performance was measured using two objective measures: task completion time and the number of input control actions. Analysis of variance and correlational procedures were used to interpret these measures. A set of 29 bipolar semantic differentials were used to form a subjective measure of consistency. The linear sum of the scores on a subset of these items was used to create a composite measure of consistency. An analysis of variance procedure was performed on the composite measure, called the Preference Index. Results show that time and subjective measures are not identical in their ability to discriminate between inconsistent versions of the interface. It is concluded that the inconsistency of the interface has different effects on subject’s ability to complete tasks as compared to subject’s ability to rate interface consistency. Interface designers should be aware that subjective ratings of interface consistency need to be collected in concert with objective performance measures to assess the effects of graphical user interface inconsistencies upon human performance. / Master of Science

As the use of computers in the workplace becomes more commonplace, the levels of interconnection and interoperability increases. The desire to pass large amounts of data almost instantaneously is the basis of the high speed local area network (LAN). To meet the needs of these LANs, the American National Standards Institute has developed the Fiber Distributed Data Interface (FDDI) standard. This new LAN can provides high speed fiber optic based communication between computers. In meeting the computer/LAN interface requirements, new methods for data transfer will be required. Trade-offs between the reliability, architecture and buffer sizes must be developed. These concepts must include variables of data transfer widths, protocol processing, transfer architecture and packet length distributions. This dissertation addresses these hardware requirements in using the high speed computer interface known as the Fiber Data Distributed Interface. / Ph. D.