HOME COMPANY PORTFOLIO RESEARCH PROMO VECTOR

 

Bluel Research

Research is not just an important part of the business plan at Bluel – it is a passion. We are continually investigating new technologies and searching for opportunities to innovate. In addition to their commercial product potential, research projects frequently produce software that can be leveraged immediately for today's applications. They also provide a useful testbed for evaluating cutting-edge development tools and methodologies.

 

KX6 Demo

The KX6 project showcases an experimental boosting algorithm for machine learning. The hexapod robot is constructed from laser-cut polycarbonate resin thermoplastic, and features an onboard microcontroller and camera with a radio link to the PC base station. The vision system uses Intel's OpenCV library for computer vision.

video
KX6_TripodGait.mov (4.16 MB)

 

Genetic Programming for Inner Loop Optimization

In a recent partnership with Tarter Econometrics, we gained access to the Pleiades distributed computing cluster. In addition to stochastic optimization applications for financial modeling, we have also developed a class of genetic programming techniques for algorithm optimization, with emphasis on numerical applications for embedded systems. Results of these and other collaborative research products conducted on the cluster will be published in upcoming technical reports.

>> Introduction to distributed high performance computing

 

Sonic Script

SonicScript™ is a Java-based multiplatform standard for audio-enhanced books. Unlike a movie, it leverages the reader's imagination while enhancing the experience with ambient sound effects and music. A working prototype has been completed for PC's, with planned extensions for portable devices such as iPod and Android. This system also includes an authoring tool called SonicScribe™.

Spam Classifier

Bumper is a new "content neutral" classifier technology for server-side filtering of spam inspired by our work with text processing algorithms. We originally approached this problem from the typical Bayesian approach; however, it quickly became apparent that this can always be defeated by a clever adversary. The key insight is that unwanted e-mail should be identified not by the content of the message, but rather by the "envelope". This motivated a new, highly-effective strategy called "bumping", which is similar in spirit to the recently proposed Sender Policy Framework ideas, except that it does not require widespread adoption of new protocols.

Bluel recently announced that the Exim-based reference implementation of Bumper will be published under an open source license. For more information on this interesting technology, click here.

 

Copyright (c) 1997-2009, Bluel Technologies Corporation