Applied Software Engineering Research Group (ASER)

Products and Services

Over the last 10 years we have developed a number of agent projects, and an agent framework called the Oak Ridge Mobile Agent Community (ORMAC). This framework facilitates the use of mobile agents that can work on one machine, then move to another machine to complete the work if needed. All messages to this agent are rerouted to new location. The framework is written entirely in Java and is portable across several system architectures.

We have developed agent-based solutions in Intelligence Analysis, Cyber Security, Geospatial Analysis, Supply Chain Management, Lean Manufacturing, Scientific Data Management, Data Fusion, and Semantic Web Applications. Our clients include: The US Army, DARPA, US Pacific Command, US 6th Fleet, Defense Logistic Agency, Intelligence Community, Lockheed Martin, Battelle Memorial Institute, and the Department of Energy.

Major Services

Research and development into challenging and unsolved problems. We typically develop proof of principle software or conduct feasibility studies in conjunction with the customer.
Collaboration and consulting on knowledge discovery issues. We sit on expert panels and committees in a variety of computer science areas.
Adapting existing tools into new environments. We work with licensees of our software to adapt them to a new environments and train their staff to maintain and support the software.

Capabilities

Intelligent Software Agents

The advancement of computing power and networking has led to the development of peer-to-peer networks and cluster computers, which are rapidly becoming commodity-type resources rather than exclusive research-type tools. To leverage these computing resources, ASER has researched, developed, and utilized a framework for the use of intelligent software agent technology in conquering difficult tasks. This framework, known as the Oak Ridge Multi-Agent Community (ORMAC), provides ASER researchers the ability to develop quickly highly refined and advanced prototype software systems for data analysis problems. Software agents developed using this framework can communicate with other ORMAC agents and have the mobility to move across different machines.

High Performance Computing

To develop robust and scalable intelligent agent architectures, the ASER group has full access to ORNL’s largest cluster computer. This machine provides up to 1.7 TFLOPS performance, 270 GB of main memory, and 11.3 TB of storage. Utilizing this resource and our ORMAC framework, ASER can provide high speed, high volume processing of data. This unique capability sets us apart from most other research organizations, and provides a value-added service to our sponsors.

Advanced Clustering Algorithms

Many of our sponsors face the challenge of understanding large amounts of data. One approach to understanding data is through a cluster analysis, which provides of means for identifying similar and dissimilar data within a data set. Unlike most clustering algorithms, which require a static data set to work, our approach to text clustering allows data to stream into the system as the data becomes available. This gives the analysts the ability to incorporate new information immediately without re-clustering the entire data set. In addition, our clustering algorithms were developed to work in a distributed manner using our ORMAC framework, allowing our sponsors to leverage the power of multiple machines working together in either a peer-to-peer environment or a clustering computing environment.

Advanced Text Analysis

With the explosion of the Internet and the resources it provides, simple text searching no longer meet the needs of our sponsors, who often require accurate analysis in a short amount of time. At ASER, we have developed new and novel techniques to address the challenges posed by text data. These techniques range from disambiguation to identifying duplicate data to extraction of entity information such as proper names of people and locations. As a specific example, we are currently researching and developing the necessary algorithms to disambiguate authors between different publications from MEDLINE. Furthermore, we are constantly pushing the speed limit of advanced text analysis to develop even faster techniques to handle even larger amounts of data.

Distributed Indexing and Information Retrieval

The demand for advanced text analysis is often coupled with the requirement for high-speed indexing and information retrieval. Many applications require efficient indexing and searching capabilities over hundreds of terabytes of data. When the data are frequently modified, faithfully reflecting changes in the search results becomes a challenging task. Existing methodologies used by modern Internet search engines can index enormous amount of information. However, the primary goal of these methods is to maximize the search throughput, but not the accuracy of the results. In other words, Internet search engines strive to service hundreds, or may be thousands, of queries per second, but the link information may not be up-to-date. These techniques cannot suffice where the accuracy of real-time information is crucial. Desktop management products, such as Google Desktop and Microsoft Desktop, on the other hand, keep track of changes made in local file systems and the search results manifest these changes. The drawback, however, is that they are not ready to handle hundreds of terabytes of data from distributed input sources. We developed a new data encoding scheme and a novel dimensionality reduction technique tailored for dynamic data. We can rapidly index large volume of data and accurately generate search results in a distributed manner.

Information Fusion

Today, it is common to have video, text, audio, and sensor data available readily on a moment’s notice. Unfortunately, this can quickly become a barrage of data capable of overwhelming even the most experienced analyst. Much of this disparate data must be fused to provide “actionable knowledge” to the human. ASER is continuing to expand the frontier of information fusion capabilities via intelligent software agent technology. We have successfully developed a variety of multi-agent systems to extract information from various sources, fuse and analyze this information, and provide higher-level, more useful results to a human.

Evolutionary Computing

We have experience in researching and developing Evolutionary Algorithms (EA’s) and Swarm Intelligence (SI) algorithm for a variety of text analysis problems. Many of today’s data analysis problems challenge the ability of traditional analysis techniques. EA’s and SI are new, non-traditional techniques that have been successfully applied to a wide variety of problem domains. The ASER group has leveraged these approaches to develop new means of clustering, sampling, and analyzing text data.

Technology

Our experience and capabilities have led to a variety of systems developed for difficult problems. In cooperation with our sponsors, we have developed the following technologies:

Piranha

We have pioneered an agent approach to text clustering that uses a large number of agents distributed over very large cluster supercomputer. This method works much faster than traditional approaches and provides the capability to cluster massive amounts of textual information in relatively short amounts of time, due to the scalability of the agent architecture. In fact, we have successfully developed this approach to work on ORNL’s largest cluster computer. For smaller date sets, Piranha is fully capable of running on a laptop system. Learn more about Piranha...

CIPHER

Professional computer attackers deliberately attempt to hide their attacks under a mass of computer network data. Working with ORNL cyber-security specialists, ASER has developed a multi-agent, information fusion system called CIPHER that fuses information from various databases to identify specifically research personnel in sensitive research areas that may be under attack from outside the ORNL computer network. This system successful scoured through over 1 million daily records of network traffic data to identify professionally implemented and targeted cyber-attacks on ORNL researchers. Learn more about CIPHER...

Image to Intelligence Archive (I2IA)

The I2IA system is a novel intelligent agent architecture that combines innovative approaches from three distinct research areas: software agents, georeferenced data modeling, and content-based image retrieval (CBIR). The overall system architecture is based on a multi-agent paradigm where agents autonomously search for images over the Internet, then convert the images to a vector used for use in searching and retrieval. This system provides the analyst with the ability to search for critical images containing features of interest. The primary data set for this system is satellite imagery; however, other types of images can be incorporated. Learn more about I2IA...

ORION

Utilizing our ORMAC agent framework, ORION is an innovative multi-agent system that can 1) autonomously search through distributed information sources and retrieve new and updated information, 2) verify the information for temporal currency and structural consistency to maintain a dynamic data archive, 3) determine the value of a particular piece of information as it relates to other pieces of information, and 4) analyze and fuse the information from different sources and formats to provide for discovery of relationships among facts and the discovery of relationships between facts and potential threat scenarios. In addition, ORION fuses information of different types such as video and text. Learn more about ORION...