Charles River Analytics has partnered with Metron and three universities to create a machine learning platform that would support the development of adaptive software systems.
Probabilistic Representation of Intent Commitments to Ensure Software Survival, or PRINCESS, will generate data on how a system, such as an unmanned vehicle’s navigation software, should act in certain situations, Charles River said Wednesday.
The companies are pursuing PRINCESS with Harvard University, the University of Southern California and the University of Birmingham under the Defense Advanced Research Projects Agency's Building Resource Adaptive Software Systems program.
Charles River focused efforts on applying PRINCESS to the automated performance of an unmanned underwater vehicle.
“With PRINCESS, we developed a general-purpose method to take a non-adaptive software system and synthesize an adaptive system from it using program transformation,” said Avi Pfeiffer, chief scientist at Charles River Analytics.
The PRINCESS team is now tailoring the approach for general use with other types of software.