A study by the US Defense Advanced Research Projects Agency (DARPA) has shown a new theory to enable autonomous military systems to learn faster and more efficiently.
Autonomous systems learn by solving problems in a risk-averse setting in simulated environments. However, this method can take months or even years to perfect, and this makes autonomous systems vulnerable when faced with unknown situations or observations in the real world.
DARPA experts took a step back to consider the purpose of high- and low-fidelity simulations. They found that while high-fidelity simulations train autonomous systems through a rote learning approach, low-fidelity environments train systems in a more general way, making them more flexible to differences in environments.
“Modeling everything with high fidelity makes it AI [artificial intelligence] The agent is more suited to the simulation dynamics,” said Dr. Alvaro Velasquez, DARPA’s program manager for the effort. “When you go into the real world, nothing looks exactly like what you modeled/simulated. We want generalized autonomy across a variety of platforms and domains.”
Get access to the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain a competitive edge.
Company Profile – free sample
Your download email will arrive shortly
We are confident in the unique quality of our company profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the form below
Additionally, the agency believes that learning and transferring autonomy in various low-fidelity simulations can lead to a faster transfer of autonomy from simulation to reality – “perhaps even in the same day versus weeks or months with traditional approaches ».
Tests of stand-alone systems
Defense AI applications cover a range of systems from unmanned autonomous vehicles to intelligence, surveillance and reconnaissance sensor systems.
GlobalData estimates that the total artificial intelligence market will be worth $383.3 billion in 2030, implying a compound annual growth rate of 21% between 2022 and 2030.
DARPA’s effort to determine the appropriate simulation conditions for training autonomous systems for a specific purpose may open a new market venture in the field of machine learning.
DARPA’s autonomous system training program will include competitions at the end of phases 1 and 2, respectively, with the results of the first competition used to narrow the selection from six to three.
Phase 1 is 18 months and will develop sim-to-sim autonomy transfer techniques and new methods to automatically develop or improve low-fidelity models and simulations to be used for transfer.
Phase 2 is 18 months and will develop sim-to-real autonomy transfer techniques and new methods to automatically develop or improve low-fidelity models and simulations to be used for transfer. There will be two competitions within the program corresponding to the two phases of the program.
Read the original at Defence247.gr