DARPA’s new “Computer Science Study Group” (CSSG)

Posted: September 8, 2008 in 2008, Articles
Tags: , , ,


If you are looking to develop some far out advanced science  project — and the folks at the Defense Advanced Research Projects Agency have a ton from airplanes that can fly for years without landing to skeletal  putty for fractured bones – then DARPA wants you.

The military’s cutting edge research agency is accepting scientists for its Computer Science Study Group (CSSG) who’s goal is to quickly identify ideas in the field of computer science that DARPA says will provide revolutionary advances to the Department of Defense (DoD).

The CSSG has, as you might guess, a wide and varied list of projects it would like to see. Some of those include:

· Machine Learning: Machine learning is the study of computer algorithms that improve automatically through experience, typically involving systems that perform tasks associated with artificial intelligence.  DARPA is interested in techniques for improving the efficiency and effectiveness of systems via the autonomous acquisition and integration of knowledge, and exploitation of this knowledge to enable continuous self-improvement.  Potential military applications include robot locomotion, wargaming, object recognition in computer vision, speech and handwriting recognition, bioinformatics, and medical diagnosis.

Smart Surveillance Systems: DARPA is interested in smart surveillance systems that use automatic image understanding techniques to extract information from the surveillance data.  In addition to proposals which consider the information extraction aspect of the challenge, DARPA will also consider those that address the use of extracted information in the context of search, retrieval, data management and investigation.

·          Bio-inspired Exploitation Systems: Bat sonar, ant colonies, and immune systems are examples of biological systems that have inspired the development of algorithms applicable to difficult and large problems in a variety of areas.  Examples include genetic and evolutionary algorithms, neural networks, new ideas for developing routing algorithms in wireless networks inspired by biology, including software and algorithms endowed with capabilities such as adaptation, evolution, growth, healing, replication and learning.  Potential applications of interest to the military include autonomous intelligent vehicles, adaptive video processing algorithms, flight and other control systems, and medical data analysis.

·          Biometrics: DARPA is interested in the development of novel and improved technologies for measuring and analyzing human body characteristics, such as fingerprints, eye retinas and irises, voice patterns, facial patterns and hand measurements, for authentication purposes.  Desirable characteristics of proposed techniques include minimizing key metrics such as the percent of invalid users who are incorrectly accepted as genuine users, the percent of valid users who are rejected as imposters, and the percent of valid users who are not recognized by the system.

·          Complexity Theory: Complexity theory deals with classifying computational problems by the amount of computational resources they require, or, more specifically, the number of processing steps and the memory required for their solution.  DARPA is particularly interested in means of determining what techniques exist for speeding up the solution of problems in high performance computing, and what the bounds on computation speed are for various types of computer architectures, including scalar, parallel, distributed network, etc.

·          Computer Vision: Computer vision is devoted to picture and video analysis to achieve results comparable to those of a human viewer.  Potential applications include medical imaging, video surveillance, detection and tracking of individuals and vehicles, and video compression.  Methods that include implementation of machine learning are of particular interest, but DARPA will also consider methods designed to solve specific tasks more effectively than previous systems.

·          Detecting Deviations from Normalcy: Pattern recognition theory tends to focus on events and patterns that are relatively constant over time.  Dynamic models of activity, however, attempt to analyze trends and extrapolate patterns to expected behavior patterns in the future.  Beyond predicting trends in patterns that an analyst might wish to detect because they represent a threat, more advanced theories might attempt to model or predict patterns that represent normal behavior, so that threats can be detected as deviations from that normalcy pattern.  Potential applications include the detection of intrusions in computer systems and networks, and the detection of medical anomalies.

·          Information Accessibility, Integration, and Management: DARPA is interested in next-generation methods, tools, and technologies to make it possible to access, integrate, analyze, and efficiently manage massive stores of widely distributed, heterogeneous information. These capabilities will help human analysts make better use of all available information resources in the pursuit of knowledge relevant to military applications.  Examples of possible research areas include development of human-computer interaction features that enable rapid, easy access to and understanding of heterogeneous information, and of cognitive systems able to “learn,” adjust to change, and repair themselves to enhance battlefield robots.

·          Network Management and Modeling: Our military services depend on a broad array of interacting physical, informational, cognitive, and social networks.  Greater fundamental network understanding is essential to insure they function reliably and smoothly, and are not vulnerable to attack.  This gap between what is known and what is needed to ensure the reliable and secure operation of complex networks makes the transition to network-centric operations problematic.  DARPA is interested in developing the fundamental knowledge necessary to design large, complex networks in a predictable manner.

·          Pattern Recognition: Pattern recognition aims to classify data (patterns) based on either a priori knowledge or on statistical information extracted from the patterns.  The patterns to be classified are usually groups of measurements or observations, defining points in an appropriate multidimensional space.  New and innovative breakthroughs in pattern recognition would be immediately applicable to information analysis.

·          Software Engineering: The process of software development and evolution is an ambitious undertaking involving complex, incomplete, sometimes inconsistent and often fuzzy factors.  Variables concerning design, quality, reliability, stakeholder interests and objectives, moving targets, and constraints such as budget and timeline must all be considered throughout a dynamic life cycle.  The challenge is to provide sound methodological support for enabling good decisions about processes and products, risks and bottlenecks as well as for selection of tools, methods and techniques.

DARPA estimates up to 12 researchers will be accepted as part of the 2009 CSSG.  August 11, 2008 is the cutoff date.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s