A Comparison of Genetic Algorithms using Super Mario Bros.
There are multiple variations of genetic algorithms (GA) that can produce effective artificial intelligence, including Evolving Behavior Trees (EBT) and NeuroEvolution of Augmenting Topologies (NEAT). The question this MQP sought to answer is: How do they compare? Using Super Mario Bros. as a benchmark, the students compared EBTs with NEAT, a GA for the evolution of Artificial Neural Networks. The results showed that NEAT had a slightly higher maximum fitness while performing poorly in all other comparisons. EBTs performed strongly in rise time, evolution time, generalization, and complexity.
Enhancing Google Docs for Essays and Peer Review
In recent years, Google has started to play a more significant role in education. In fact, more than 40 million students, teachers, and administrators are active users of Google Apps for Education. Those users can take advantage of Classroom, which is a way to organize assignments, provide feedback, and enhance class communication. However, The current version of Google Docs lacks some features that would be helpful to teachers grading any free response style question. Comments must be entered manually, and can be resolved without any explanation, meaning that students learn less from peer comments. Our goal in this project is to enhance Google Docs with an add-on that will help teachers grade essays more efficiently while also improving the peer review process. We wanted to allow teachers to generate comments quickly and be able to summarize them with the click of a button.
Visualizing Graphs of Tracks
BAE systems, a multi-national defense company, must interpret and visualize massive quantities of geo-temporal track data. While BAE currently has some tools that are capable of representing this data, existing tools cannot effectively represent association likelihoods between track entities. This project provides a software solution called TrackWiz for interacting with and visualizing this complex data and for interaction track associations.
Wearable Action Guidance Bands
The Wearable Action Guidance (WAG) System is a training tool designed to improve the efficiency and convenience of teaching and learning new physical skills, while matching or exceeding the quality of feedback received from an in-person trainer. The system consists of a computer application and a set of wearable bands; each band comprises a 3D printed case, an inertial measurement unit, a battery, a ring of vibration motors, and a secure strap. Trainers can use the WAG System to record and save motions for distribution to trainees, while trainees can use the system to play back those motions with directed vibratory feedback. Initial prototypes have attracted potential partners interested in introducing the technology to various markets including athletic training and physical rehabilitation.
Graphically Represented Image Processing Engine (GRIP)
The goal of this project was to build an application that could be used to construct and deploy computer vision algorithms. Developing a vision program can be difficult because it is hard to visualize the intermediate results. Java and OpenCV were used to implement a graphical development tool. This simplifies and accelerates the creation of vision systems for experienced users and reduces the barrier to entry for inexperienced users. As a result, many teams with minimum computer vision knowledge successfully used our software in the 2016 FIRST Robotics Competition game.