WHAT WE DO


CASE STUDIES

The scope of Knexus Research’s accomplishment is both deep and diverse, as its defense industry clients turn to us to create trailblazing solutions to their most challenging problems. 

Our award-winning team’s unparalleled intellectual and technical capabilities combine to generate novel answers to original questions through research, experimentation, innovation, and rigorous testing. Here are just a few projects we have tackled in recent years, reflecting the wide-ranging scope of our scientists’ and engineers’ expertise.

Cognitive Architecture for RePlanning & Execution (CARPE)


Sea mines are even more dangerous than before as the technology improves and provides an effective counter against expensive assets. Mine countermeasures assets and operations have also evolved in response to such threats by exploiting new sensors and unmanned vehicles, for example. Mine Countermeasures Commanders (MCMC) must make plans and respond to unexpected events in an evolving highly risky, uncertain, and dynamic operating environment. However, they lack the tools to help them efficiently and effectively plan a response to these threats, which can now take days and unexpected events may take hours to respond. This prompted the need for a next generation tactical decision aid that could help them respond to unexpected events in minutes.

 

In response, Knexus AI scientists and engineers developed a decision engine comprising a range of automated situation analysis and rescheduling algorithms that alert the MCMC staff as performance anomalies develop and immediately present them with a handful of valid options for responding to the developing situation. In a series of evaluations conducted by MCM subject matter experts the system showed that early detection of performance anomalies combined with valid rescheduling support enabled them to successfully respond in minutes. It resulted in significant reduction in operation time and reduced transistor and warfighter risk as well. Knexus published its findings in two peer reviewed conferences and filed a patent on its innovative algorithms. The project has been transitioned to a Future Naval Capabilities program to further mature the technology.

Behavior Recognition and Threat Assessment (BRETA)


The Navy is constantly pushing the frontiers of AI and machine learning approaches for improving the performance on complex task such as threat detection and high value asset protection. Knexus scientists and researchers developed new algorithms for maritime object detection and maritime activity and behavior recognition that outperformed the state-of-the art.

 

They developed proof-of-concept algorithms and systems and evaluated the effectiveness of their algorithms on realistic maritime traffic data. They reported their finding in three peer reviewed articles.

Spatio-temporal Cognition for Language and Scene Understanding


Language understanding and generation is closely tied to visual and other sensory perceptions, subsequent conception, and action. However, despite the abundance of theories, none seem to provide an integrated computational framework at the time we conducted this research. Knexus scientists and engineers developed an integrated theory of representation, learning and reasoning to explain the connection between perception and language.

 

They focused their efforts on generating representation of English spatial terms, such as on, in, about, over in several computational tasks such as object placement and locative question answering. The evaluation of their models and theories in virtual 3D office environments demonstrated the validity and the success of their approach They published two peer reviewed articles and filed three patents, one of which has been awarded. They transitioned their research findings into applications that enables language based command and control of simulated autonomous ground vehicles.

Intelligence Question Answering System (INQA)


The Office of the Director of Intelligence (ODNI) entrusted IARPA to investigate approaches for next generation of question answering(QA) technologies to support the task of Intelligence Analysts.  Five teams from leading universities and research labs pushed the frontiers for technology to enable users to ask question in natural language and receive precise answers in response. Knexus scientists and engineers developed a framework to integrate, showcase and the power of these algorithms for the intelligence analysts.

 

The Knexus team helped design the experiments and conduct the associated human subject studies.  The Knexus team received accolades from the program regarding timely delivery of system and results for a highly complex system.  Their prototype system was selected for advanced technology demonstration for the U.S. Congress at the Capitol Hill.