Research Areas

Robot dogs in action

Autonomous Cognitive Mobile Robots in Dynamic Environments
This research, lead by Mohan Sridharan, aims to enable mobile robots to operate autonomously in dynamic environments, interacting with humans and other robots. The mobile robots will fuse the input from multiple sensors in order to learn environmental models, detect environmental changes, and adapt the learned models in response to such changes. Furthermore, the robots will learn from, and collaborate with humans by effectively binding the information obtained through different modes of interaction. Specifically, algorithms will be devised to enable mobile robots to:

Learn appropriate environmental models based on sensory input and, to use the learned models to incrementally refine each other (i.e. bootstrap) in order to detect and adapt to environmental changes.
Bind information across different modes of interaction (visual, speech etc) and, to probabilistically plan actions using sequential decision processes, e.g. Partially Observable Markov Decision Processes (POMDPs), in order to achieve the desired goals through an optimal use of available resources.
Fuse the information from several noisy sensors to exploit the benefits of their mutually complementary properties, in order to learn accurate environmental models.

Furthermore, a heterogeneous team of such learning and adapting robots will be enabled to take collaborative action in the presence of noise in sensing and communication. The mobile robots will be deployed in specific application testbeds that require real-time performance in dynamic environments, both outdoors and in the context of the Robot Soccer Initiative, in order to learn abstractions that generalize across potential applications. For more information on current (and previous) research and publications, including images and videos of robots in action, click here.

Data miners at work.

Classification Rule Induction from Databases
This research, lead by Dr. Rattikorn Hewett, develops data mining techniques and algorithms for extracting information from structured databases. The project includes ongoing development of the SORCER(Second-Order Relation Compression for Extraction of Rules) system. SORCER is based on our induction algorithm referred to as table compression.

Mu-Analysis.

Mu-Analysis (Mutation Analysis)
This research is lead by Dr. Akbar Siami Namin.

Mutation is the process of transforming elements of a piece of code into another set of elements in the same class using a set of functions called mutation operators. Mutation testing is an adequacy measurement for a set of test cases. Recently, application of mutation in empirical studies has been of interest. Studies show that mutation is an effective tool in assessing testing techniques. However, there exist several challenging issues regarding its definition and feasibilities. Inconsistency in defining mutation operators, infeasibility of using mutation due to its high cost, and insufficient analysis on mutation and its characteristics are a few issues that we investigate in this project. Generally, the project aims at different aspects of mutation:

Definition
Analysis
Feasibility
Application

Related keywords: Software Testing, Mutation Testing, Mutation Analysis, Empirical Study

TrustTTU.

TrustTTU (TRUST managemenT in disTribued compUting)
This research is lead by Dr. Akbar Siami Namin.

TrustTTU aims at tackling the challenging issues with trust and reputation in heterogeneous environments. Entities need to collaborate with unknown entities and share the resources. However, the challenging question is: how to trust an unknown entity whose service is in demand. We have applied several cryptographic methods such as secret sharing schemas, metering schemas, and hash functions in order to model the trust issues in secure and invulnerable manners. This project mainly focuses on the following issues of trust management:

Definition
Security
Decision Making
Modeling

Related keywords: Internet, Trust Management, Reputation, Security, Cryptography

AgentSOA.

AgentSOA (cooperAtive intelliGENT reSOurce shAring)
This research is lead by Dr. Akbar Siami Namin.

This project models cooperative environments by integrating intelligent software agents and service-oriented architecture. Both technologies have advantages and pitfalls. Intelligent agents are proactive components using their ability in decision making. However, agents are tightly coupled and they communicate in a limited way. In contrast, in service-oriented architecture components are independent and loosely coupled. Services-oriented component, however, are less or no proactive compared to software agents. This project models integration of these two technologies with respect to:

Meta-Modeling
Resource Sharing
Grid Computing

Related keywords: Internet, Resource Sharing, Intelligent Software Agent, Service-oriented Architecture

Computer Science at Abilene

Research Areas