The Defense Advanced Research Projects Agency has introduced a research program to explore methods to increase the predictability and safety of learning-enabled autonomous platforms such as unmanned aerial and driverless vehicles.
DARPA said Wednesday researchers will work to create tools that reinforce trustworthiness at the design stage and address functional and safety goals explicitly stated under the Assured Autonomy program.
“Tremendous advances have been made in the last decade in constructing autonomy systems, as evidenced by the proliferation of a variety of unmanned vehicles,” said Sandeep Neema, a DARPA program manager.
“In spite of these advances, deployment and broader adoption of such systems in safety-critical [Defense Department] applications remains challenging and controversial,” Neema added.
DARPA aims to integrate the program’s resulting technology into learning-enabled, cyber-physical systems design tool chains for use in the defense and commercial sectors.