So the robot creates a hypothetical state, then runs a simulation onit From: [email protected] To: [email protected] Subject: [agi] clever new approach to robotic scene understanding Date: Sat, 22 Jun 2013 12:58:05 +0100
This is pretty smart: ------------------ Robots learn how to arrange objects by 'hallucinating' humans into their environment (w/ video) June 21st, 2013 in Electronics / Robotics Credit: Cornell University (Phys.org) -A team of robotics engineers working in the Personal Robotics Lab at Cornell University (led by Ashutosh Saxena) has developed a new way to give robots a context-sensitive way to organize a room. Instead of providing the robots with a map, the researchers instead cause the robot to "imagine" how a human being would use objects in a room and then to place them accordingly. Traditional programming for robots has relied on providing clear instructions on what they are supposed to do-pick something up from one place and put it down in another, for example. To get a robot to engage in activities that require some degree of intuition, however, would mean giving them some means for doing so. One example would be to ask a robot to enter a room, note objects on a table, and ask that they be arranged on a desk for use by a person. To arrange objects in a way that makes sense to a human being requires some understanding of how people operate. To do that, the Cornell team gave a test robot a means for imagining what a person would look like in the room while using a set of objects. As an example, the researchers programmed a robot to pick up a coffee mug and computer mouse from a table and place them on a desk in what would seem the most logical positions based on human behavior. To do that, they gave the robot what they call "an ability to hallucinate" humans into the room-the robot "brain" overlays images of stick figure humans onto images of the room. Various poses are considered while the robot "imagines"; how a human might make the best use of the mouse and mug. Based on this process, the robot then placed the mouse just to the right of the keyboard (because the average person is right handed) and the mug a little ways back-within reach, but not so close it might get knocked over unintentionally. The approach mimics what a human would likely do given the same instructions, of course, and that is exactly the point. [Video available on website] Hallucinating humans for robotic scene understanding. The team will be outlining their research findings and progress at the Robotics: Science and Systems 2013 conference in Berlin next week. More information: Project page: pr.cs.cornell.edu/hallucinatinghumans/ Research papers: Infinite Latent Conditional Random Fields for Modeling Environments through Humans (PDF) Hallucinated Humans as the Hidden Context for Labeling 3D Scenes (PDF) via IEEESpectrum © 2013 Phys.org "Robots learn how to arrange objects by 'hallucinating' humans into their environment (w/ video)." June 21st, 2013. http://phys.org/news/2013-06-robots-hallucinating-humans-environment-video.html ------------------ Or as the researchers themselves put it: “For scene understanding, one popular approach has been to model the object-object relationships. In this paper, we hypothesize that such relationships are only an artifact of certain hidden factors, such as humans. For example, the objects, monitor and keyboard, are strongly spatially correlated only because a human types on the keyboard while watching the monitor. Our goal is to learn this hidden human context (i.e., the human-object relationships), and also use it as a cue for labeling the scenes. We present Infinite Factored Topic Model (IFTM), where we consider a scene as being generated from two types of topics: human configurations and human-object relationships. This enables our algorithm to hallucinate the possible configurations of the humans in the scene parsimoniously” http://www.cs.cornell.edu/~asaxena/papers/jiang-hallucinatinghumans-labeling3d-cvpr13.pdf ____ Looking at objects and scnes in terms of how humans use and relate to them, is consistent with the Barsalou idea of concepts as simulators of how we relate to objects and scenes. I of course v. much like these guys’ use of graphics/stick figs.. The limitations of their algorithmic approach, though, are that they can only use a limited set of figure poses, whereas the human mind’s figures are fluid and more or less infinitely reconfigurable, re-posable. AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
