Organizational Unit:
Undergraduate Research Opportunities Program
Undergraduate Research Opportunities Program
Permanent Link
Research Organization Registry ID
Description
Previous Names
Parent Organization
Parent Organization
Organizational Unit
Includes Organization(s)
ArchiveSpace Name Record
Publication Search Results
Now showing
1 - 8 of 8
-
ItemUsing First Order Inductive Learning as an Alternative to a Simulator in a Game Artificial Intelligence(Georgia Institute of Technology, 2009-05-04) Long, Kathryn AnnaCurrently many game artificial intelligences attempt to determine their next moves by using a simulator to predict the effect of actions in the world. However, writing such a simulator is time-consuming, and the simulator must be changed substantially whenever a detail in the game design is modified. As such, this research project set out to determine if a version of the first order inductive learning algorithm could be used to learn rules that could then be used in place of a simulator. By eliminating the need to write a simulator for each game by hand, the entire Darmok 2 project could more easily adapt to additional real-time strategy games. Over time, Darmok 2 would also be able to provide better competition for human players by training the artificial intelligences to play against the style of a specific player. Most importantly, Darmok 2 might also be able to create a general solution for creating game artificial intelligences, which could save game development companies a substantial amount of money, time, and effort.
-
ItemDistributed Feature Extraction Using Cloud Computing Resources(Georgia Institute of Technology, 2009-05-04) Dalton, StevenThe need to expand the computational resources in a massive surveillance network is clear but traditional means of purchasing new equipment for short-term tasks every year is wasteful. In this work I will provide evidence in support of utilizing a cloud computing infrastructure to perform computationally intensive feature extraction tasks on data streams. Efficient off-loading of computational tasks to cloud resources will require a minimization of the time needed to expand the cloud resources, an efficient model of communication and a study of the interplay between the in-network computational resources and remote resources in the cloud. This report provides strong evidence that the use of cloud computing resources in a near real-time distributed sensor network surveillance system, ASAP, is feasible. A face detection web service operating on an Amazon EC2 instance is shown to provide processing of 10-15 frames per second.
-
ItemPerformance Information Sharing Middleware(Georgia Institute of Technology, 2008-05-05) Reiss, CharlesThis thesis presents a design for distributed monitoring system designed to enable monitoring-informed optimizationsin distributed applications. Microbenchmarks and an evaluation in a scientific-computing scenario are presented. The monitoring system is intended to assist when application requirements cannot be easily expressed in a form suitable for existing autonomic computing approaches. The design embeds awareness of the application's topology into the monitoring system so queries can reference a node's place in the application without embedding extra assumptions about the overall layout of the application. Through integration with dynamic code generation, users may make potentially application-specific metadata available and use such data within dynamically deployed filters and transformation functions. Evaluations demonstrate that this approach can provide timely and useful information with low overhead.
-
ItemLearning Behaviors Through Demonstration: Artificial Intelligence for Non-Player Characters in an Interactive Drama(Georgia Institute of Technology, 2007-12-17) Amundsen, Thomas CharlesIn both the game industry and the academic community, there have not been many attempts to create complete interactive drama systems. I am working with a group that will be the first to undertake this daunting challenge, and one of the most important components in such a system is artificial intelligence (AI) to control the behaviors of non-player characters. Traditional approaches to designing the AI for any kind of interactive game involves designing characters who follow scripted behaviors. This method is cumbersome and amounts to gameplay that is repetitive and thus inhuman. This paper will describe my attempt to design a system that will allow an expert user to demonstrate behaviors to the system, which the system will use to learn how to behave on its own using case-based reasoning. The end result of this work will not only benefit interactive dramas but may help in the design of game AI for other genres of video games or simulations.
-
ItemEvaluating the effectiveness of using touch sensor capacitors as an input device for a wrist watch computer(Georgia Institute of Technology, 2007-12-17) Wilson, GregoryOn the go computing is becoming more important for users who wish to access information from anywhere. Wearable computers are an optimal solution to achieving this feat because it allows for easy accessibility and quick use. There are many challenges that arise with small computers worn on the body. One of the most common issues is the interaction between the computer and the user and more specifically how the user enters input. In this paper we research a potential effective way to interact with a wrist watch by mounting touch sensors on the watch band.
-
ItemJacket’s Garage: Software for a Physical Science Learning Environment(Georgia Institute of Technology, 2007-05) Kramer, Christopher W. ; Kolodner, Janet L.The Learning by Design lab has been developing software and curriculum to support learning in the context of design challenges. We have used software to guide learners through design and investigation practices as well as help them describe their scientific observations. In this paper, we present easy-to-use software tools that support learners in their construction of robust science conceptions. We investigate how multiple representations such as diagrams, graphs, and animations can provide support for knowledge construction. In addition, we investigate how explanation templates scaffold learners’ interpretations of science. We have combined these tools into one software environment, Jacket’s Garage.
-
ItemAn Integrated OS- plus VMM-bypass Solution for Virtualized I/O(Georgia Institute of Technology, 2007-05) Kurjanowicz, Matthew DavidCurrent Remote Direct Memory Access (RDMA) technologies are either not virtualization aware or run on proprietary connection fabrics such as InfiniBand, Myrinet and Quadrics. The introduction of 10-Gigabit Ethernet provides a high-speed connection that utilizes existing network infrastructure. As the first step to provide a useful, virtualization-aware RDMA implementation we propose an integrated OS- and VMM-bypass solution implemented on the Netronome NFE-i8000 Network Processor utilizing the Intel IXP 2855. We use this implementation to demonstrate the feasibility and usefulness of an integrated VMM- and OS-bypass communications engines that can be used to implement such a virtualized RDMA solution and to understand the architectural limitations of such an implementation.
-
ItemA Dynamic Approach to Statistical Debugging: Building Program Specific Models with Neural Networks(Georgia Institute of Technology, 2007-05) Wood, MatthewComputer software is constantly increasing in complexity; this requires more developer time, effort, and knowledge in order to correct bugs inevitably occurring in software production. Eventually, increases in complexity and size will make manually correcting programmatic errors impractical. Thus, there is a need for automated software-debugging tools that can reduce the time and effort required by the developer. The performance of previously developed debugging techniques can be greatly improved by combining them with machine-learning. Our research focuses on the application of neural networks within the domain of statistical debugging. Specifically, we develop methods to mine statistical debugging data that can then be used to train neural networks; these generated multi-layered neural networks can then be used to identify suspicious programmatic entities. Our developed networks are generated on a per program basis in order to leverage specific programmatic properties. In our empirical evaluation we compare our proposed approach with a state-of-the-art automated debugging technique. The results of the evaluation indicate that, for the cases considered, our approach is more effective than the considered technique.