Closing the perception-actuation loop using machine learning: New perspectives and strategies
Director of Robotics,
Click on any part of the transcript to go to that point in the video.
Recent advances in perception technology, fueled by progress in deep learning, have materially changed the degree of situational awareness one can expect from robots engaged in the real world: in addition to perceiving the geometry of the world around them, robots can now also reason about its semantics and communicate intuitively with the people sharing their environment.
Yet, we're arguably still struggling to deploy robots in human-centered environments. Much of the difficulty centers around closing the loop between perception and actuation in a manner that's safe, reliable, precise, and flexible. This talk explores recent progress in machine learning which directly addresses these challenges and opens up new avenues in connecting perception and behaviors in real-world environments.