CS Opportunities

Simulating Motion Plans for Large Robot Teams

by Srinivas Akella

We are developing algorithms to coordinate the motions of large numbers of robots. This NSF-funded project will involve visualizing the motions of robots in 2D and 3D using descriptions of their motion plans (for example, like the Intel drone light show at the 2017 Super Bowl halftime show, www.intel.com/content/www/us/en/technology-innovation/aerial-technology-…). Students will get to experiment with new algorithms for motion planning of hundreds of robots moving from one formation to another goal formation, while ensuring there are no collisions between the robots. Students will be encouraged to develop simulations and potentially even perform experiments on real robots. We are looking for students with strong programming skills (must be comfortable creating 2D/3D visualizations in at least one of MATLAB, Python, or C++ with the OpenGL library) and with a good background in Mathematics (especially trigonometry, basic linear algebra, and calculus).

Internet of Things (IoT) and Active Learning

by Mohsen Dorodchi

Internet of Things (IoT) is an emerging area in computing. It is expected to offer promising solutions to transform the operation and role of many existing systems. This research project focuses on IoT in Active Learning classrooms to help with the teaching and learning. In particular, the focus is on identifying the correlation of the designed activities and students’ activeness who are working in small groups in an activity-based active learning classroom.

Untitled Project

by Siddharth Krishnan

How do memes spread on Facebook? How and when does a hashtag become popular? Can we forecast/predict viral content? How can we harness information cascades to make ‘real-world’ predictions? I am broadly interested in web-mining, data analytics, computational social science, and applied machine learning with a primary emphasis on analyzing, characterizing, and forecasting information (news, rumors, memes, advertisements, etc.) dynamics on online social networks & social media. Furthermore, my research aims to leverage dynamical processes (like cascade propagation) to build explanatory & predictive models of actions of large groups of people and societies.

Pre-training Synthetic Gradient Descent

by Minwoo Jake Lee

In deep networks, synthetic gradient descent approximates the gradient with separate networks. This decoupled networks can be beneficial since it can remove the dependance between each node, and enable the parallel computation. This project examines the feasibility of pertaining the synthetic gradient descent networks.

Reading Deep Networks with Hidden Markov Model

by Minwoo Jake Lee

Readability and analysis of deep neural networks are common problems in deep learning study. A few studies start to suggest a way to visualize or read the learned networks for interpretation of their solution. Here, we examine state transition model analysis using hidden Markov model for the networks.

Analytical Reinforcement Learning with Sparse Bayesian Model

by Minwoo Jake Lee

Reinforcement learning provides a way to train an intelligent agent to learn or develop a policy based on the rewards from the environment. This requires a significant amount of experience to find a decent policy. From the experience, it is difficult to analyze the solution and domain. By using the Sparse Bayesian Reinforcement Learning model, we try to answer the questions as follows. What is core factor to make the policy work? What was the main reason that caused the error on the domain or system? Is there any feature we neglected for the learning model design? Is there possible improvement of solution?

Practice for Intelligent Agent

by Minwoo Jake Lee

Problem solving requires the presentation of it. Machine learning and reinforcement learning assume that the target problem to tackle is given in many cases. However, some machine learning studies such as deep learning and self-taught learning have shown that learning without the actual target problem can help learning the target problem. Practice suggests a way that we can make intelligent agent gain meaningful information before solving the target problem. We expect that this approach can reduce the cost and time for tackling the target problem and also further improve the performance of learning.

Immersive Visualization and Interaction

by Aidong Lu

This project is to explore the latest virtual and augmented reality devices and design the next generation of immersive analytics systems that allow users to mix data and the real physical world. These new visual and interactive interfaces open opportunities for creating multi-sensory experiences with touch, voices, gestures, etc. that are significantly different than the traditional keyboard and mouse.

Human Behavior Modeling

by Samira Shaikh

We are interested in detecting and modeling all aspects of human behavior, especially those that can be inferred from language and natural language communication.

Intelligent Agents/Chatbots

by Samira Shaikh

This research explores the development and use of intelligent agents and chat bots that can model aspects of human behavior, for example persuasive behavior, in social interaction.

Emotion Detection from Text

by Samira Shaikh

This project will explore the computational detection and modeling of emotion from text, for eg. social media. The project will provide opportunities to learn techniques in natural language processing and applying deep learning techniques to large bodies of text.

Predictive “Keyboard” for People with Disabilities

by Min C. Shin

We all have benefited greatly from predictive keyboards on our devices. They take our imperfect typings into understandable sentences. Could we bring such benefit to people with physical disabilities? Rather than extremely slow “typing” through gaze to simply say “I -a-m h-u-n-g-r-y,” can we combine the visual understanding as well as context to translate a simple gaze at dining room around noon time to trigger “I am hungry?” Then allow the user to train the algorithm to continually learn through a simple feedback of correct or incorrect?

Perceptual studies of Shapes for Visualization Design

by Kalpathi Subramanian

This ongoing project focuses on studies of symbols (shapes) for use in information visualization, specifically scatterplots. A number of studies have been completed with promising results. The student will work with a senior doctoral student on designing and conducting new studies to study the application of the initial results into routine scatterplot data sets, to get a deeper understanding of the factors influencing visualization design.

Graph Visualization

by Kalpathi Subramanian, Erik Saule

The goal of this project is to look at very large graphs and algorithms to visualize them in meaningful ways, so as to emphasize graph structure. A key question to answer is how can visualization help understand graphs and their underlying properties. We will look at large graphs from a number of domains, including social networks, street networks, etc.

Interactive Textbook Project

by Kalpathi Subramanian, Erik Saule

This newly started project will focus on building interactive modules targeted freshmen/sophomore level courses in Computer Science. The goal is to make these modules engaging, interactive and visual, while also using sophisticated methods to maintain academic integrity and quality. The current project targets modules in data structures and algorithms and use existing tools and repositories to promote wite dissemination.

Aerial Multimedia Networking of Swarming Drones

by Pu Wang

Mobile aerial multimedia sensing can provide timely, adaptive, and enriched multimedia observation of harsh, hostile, or remote environments through deploying a swarm of mobile aerial sensing platforms, i.e., quadcopters equipped with cameras, mircophones, scalar sensors, autopilot systems and wireless modules. Mobile aerial multimedia sensing promises many emerging applications, such as disaster assessment (such as flooded Houston by Hurricane Harvey), remote structural health monitoring, pollution source tracking, intelligent agriculture, and autonomous surveillance and patrol. One of the key challenges faced by aerial multimedia sensing is to enable long-range, high-throughput and QoS-guaranteed multimedia communications using bandwidth-limited air-to-ground channels with highly dynamic behavior caused by inherent drone mobility along with complex multipath fading and shadowing. To counter this challenge, students will learn, implement and test aerial wireless communication schemes that allow the drones to intelligently change its operating frequencies, power, and modulation schemes according to surrounding environment dynamics.

Design and Development of Wireless Underwater Robots

by Pu Wang

The students will work with the PhD and Master students from Dr. Wang’s lab to develop and realize a wireless underwater robot network testbed, which consists of a group of underwater mobile platforms equipped with fine-grained sensing module, embedded AI computing device, underwater autopilot system, and high-speed underwater wireless transceivers. Such platform can pave the way for a wide spectrum of underwater applications, such as underwater automatic photography, pollutant source seeking, water quality monitoring, tsunami and seaquake warning, surveillance of leakage in underwater oil and gas pipelines, mine reconnaissance, fish survey and behavior study, and underwater search and rescue.