Computer Science Honors Papers

Visualizing the Novel

Clinton Mullins — Thu, 23 May 2013 11:21:07 PDT

Real-time Control of a Robot Arm Using an Inexpensive System for Electroencephalography Aided by Artificial Intelligence

James O'Connor — Thu, 23 May 2013 11:21:06 PDT

Indoor Scene Understanding

Bo Xiong — Thu, 23 May 2013 11:21:05 PDT

In computer vision, holistic indoor scene understanding from images is a complex and important task that requires solving several subtasks simultaneously. Advanced indoor scene understanding systems can be used to build personal robots that can clean rooms and to build automated surveillance systems to improve public safety.

In this research, we consider the problem of indoor scene understanding from both RGB and depth images. We are interested in distinguishing the major planes from objects that appear in a scene and our goal is to classify each pixel from an indoor image as one of the following categories: walls, the ceiling, the oor or objects.

We present how to construct descriptors from RGB and depth images and demonstrate that images with depth can be used to improve the performance of our system. Our recognition system is based on support vector machines and a Markov random eld. In addition, we explore how to utilize global consistency to improve the performance of our system. In particular, we propose two ltering algorithms that can improve the performance over state-of-the-art computer vision systems.

Punctuated Anytime Learning and the Xpilot-AI Combat Environment

Phillip Fritzsche — Tue, 31 May 2011 08:34:28 PDT

In this paper, research is presented on an application of Punctuated Anytime Learning with Fitness Biasing, a type of computational intelligence and evolutionary learning, for real-time learning of autonomous agents controllers in the space combat game Xpilot. Punctuated Anytime Learning was originally developed as a means of effective learning in the field of evolutionary robotics. An analysis was performed on the game environment to determine optimal environmental settings for use during learning, and Fitness Biasing is employed using this information to learn intelligent behavior for a video game agent controller in real-time. Xpilot-AI, an Xpilot add-on designed for testing learning systems, is used alongside evolutionary learning techniques to evolve optimal behavior in the background while periodic checks in normal game play are used to compensate for errors produced by running the system at a high frame rate. The resultant learned controllers are comparable to our best hand-coded Xpilot-AI agents, display complex behavior that resemble human strategies, and are capable of adapting to a changing enemy in real-time. The work presented in this paper is also general enough to further the development of Punctuated Anytime Learning in evolutionary robotic systems.