Communications of the ACM
Autonomous Subs Wayfind Without GPS
Researchers test deep reinforcement learning systems for autonomous submarines, for now, in a pool at Flinders University in Adelaide, Australia.
Credit: T. Chaffre/Flinders University
Researchers led by Thomas Chaffre at Australia's Flinders University used reinforcement learning to improve the ability of uncrewed underwater vehicles to navigate without GPS and while compensating for interference from ocean currents.
The researchers altered the reinforcement learning's memory buffer system, which stores outcomes from past actions that are sampled randomly during the training process to update the model's parameters.
The change ensures that when sampling from its memory buffer, the model gives more weight to actions that led to large positive gains and those that occurred more recently.
From IEEE Spectrum
View Full Article
Abstracts Copyright © 2023 SmithBucklin, Washington, D.C., USA
No entries found