Our staff has a 25 year history of creating physics simulations and writing deeply technical software using the technology constraints of the client’s organization (which oddly once lead to the creation of an AI system using SQL stored procedures). Highlights include:
EPA Motor Vehicle Emission Simulator (MOVES). As a contracted principal analyst for the US Environmental Protection Agency, Mr. Faler designed and programmed the MOVES and National Mobile Inventory Model (NMIM) software using Java and MySQL grid computing systems to help users develop estimates of current and future emission inventories for on-road motor vehicles and non-road equipment. A core figure of the MOVES and NMIM design efforts since 2001, he has expanded the code and database architecture for both speed and functionality as well as implemented and tested many new models for pollution calculation. Recent work includes porting MOVES to a cloud computing system, 20x speedup by converting SQL to Go, enhanced evaporation emission model development, and a complex fuel effect modeling subsystem complete with an embedded optimizing equation compiler. Participated in all public releases of the MOVES and NMIM models. Presented on cloud computing and MOVES design at EPA conferences.
Genetic Programming. A new company corrects postal addresses rejected by the United State Postal Service. Mr. Faler, along with the founder of the new company, created the core algorithms behind this process. The “Genetic Programming” (“GP”) technique was used to automatically generate models for correcting the addresses. These models were then integrated with a human generated meta-model. The GP technique is similar to the more common “Structural Equation Modeling” statistical analysis technique in that both can produce an equation describing multivariate relationships. GP, however, is able to generate not only an equation, but an entire modeling algorithm, complete with conditionality and iteration. Also used GP to reverse engineer equations contained within compiled code.
Lunar Rover Communications. The Part-Time Scientists GmbH group is planning a mission to land a rover onto the moon in 2017. To support this mission, they elected to design their own communication protocol between the rover and the array of supporting earth stations. Critical to mission success, protocol selection is a complex multiple objective optimization problem. Mr. Faler developed mission simulator software for selecting the protocol parameters. The simulator uses C++ code running on a GPU to achieve a 7 fold increase in simulation performance over other C++ and Java-based mission simulations. A parameter set with a 25% increase in mission success probability over the best parameter set selected by human committee was found. During the project, Mr. Faler also created two new artificial intelligence (AI) techniques. He was a keynote speaker at the 2012 NVidia GPU conference to highlight his pioneering work.
KickSat Sprite Software. Wrote software flown into space onboard the KickSat satellite in 2014. The software performed all operations of a Sprite satellite to be deployed by KickSat. Of special note are the software’s use of redundant memory and an exceptionally robust telemetry protocol.
Compressed Sensing. Created and presented a GPU framework for using Compressed Sensing to halve the error rate in reconstructed satellite telemetry. Created a Compressed Sensing RF signal correlator for deep space communication systems.
Heuristic Location. Created a patented sensor fusion algorithm for locating a signal using only one or two receivers instead of the classic three needed for triangulation. The algorithm is in use tracking fire fighters.