Loon’s navigation system’s most complex task--steering balloons--now is assisted by artificial intelligence, specifically deep reinforcement learning, a type of machine learning, Loon says.
Overall, the RL system kept balloons in range of the desired location more often while using less power, according to Loon.
Reinforcement learning enables an agent to learn by trial and error in an interactive environment using feedback from its own actions and experiences, Loon says. “As far as we know, this is the world’s first deployment of reinforcement learning in a production aerospace system.”
This contrasts against the conventional approach of the automated system following fixed procedures artisanally crafted by engineers, Loon notes.
In this case, RL helps optimize flight in the face of time-varying winds, partial visibility of the wind field surrounding the balloon, not always having enough power for the ideal maneuver and frequent decision points.
Google AI teams hope that beyond Loon, this work on stratospheric navigation can serve as a proof point that RL can be useful to control complicated, real world systems for fundamentally continual and dynamic activity.
No comments:
Post a Comment