ARLINGTON, Va. – U.S. military researchers are asking industry for ways of developing real-world autonomous systems quickly from models and simulations. These autonomous systems should be able to learn from their experiences to adapt to rapidly changing conditions.
Officials of the U.S. Defense Advanced Research Projects Agency (DARPA) in Arlington, Va., issued a solicitation on Tuesday (DARPA-PS-23-13) for the Transfer from Imprecise and Abstract Models to Autonomous Technologies (TIAMAT) project.
TIAMAT seeks to develop fast autonomy transfer methods that learn and refine models and simulations based on experience to enable same-day autonomy that not only is robust to quick changes in dynamic environments, but also is adaptable to a variety of military systems.
The three-year program is divided into two 18-month phases: to develop sim-to-sim autonomy transfer techniques and methods for low-fidelity models and simulations; and simulation-to-real-autonomy transfer techniques for moving this machine autonomy to real-world systems.
Generally there are two directions for improving the transfer of learned autonomy from simulation to reality: increase simulator fidelity; and develop algorithms that learn from low-fidelity simulations.
A key component of this program seeks to refine the models and simulations used for machine learning and transfer based on relevant experience collected in simulation and in the real world. This establishes a natural feedback loop between autonomy transfer and abstraction refinement for effective transfer of machine autonomy.
This approach has two primary challenges: how to conduct autonomy transfer quickly; and how to refine the models and simulations based on agent experience.
Companies interested should email abstracts to [email protected] no later than 17 Oct. 2023. Companies submitting promising abstracts may be invited to submit full proposals to the DARPA Solicitation Website at https://baa.darpa.mil no later than 20 Dec. 2023.
Email questions or concerns to DARPA at [email protected]. More information is online at https://sam.gov/opp/0399474725314fd58f34bc29849ce305/view.