eduzhai > Applied Sciences > Engineering >

Depth by Poking Learning to Estimate Depth from Self-Supervised Grasping

  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: Accurate depth estimation remains an open problem for robotic manipulation;even state of the art techniques including structured light and LiDAR sensorsfail on reflective or transparent surfaces. We address this problem by traininga neural network model to estimate depth from RGB-D images, using labels fromphysical interactions between a robot and its environment. Our networkpredicts, for each pixel in an input image, the z position that a robot s endeffector would reach if it attempted to grasp or poke at the correspondingposition. Given an autonomous grasping policy, our approach is self-supervisedas end effector position labels can be recovered through forward kinematics,without human annotation. Although gathering such physical interaction data isexpensive, it is necessary for training and routine operation of state of theart manipulation systems. Therefore, this depth estimator comes ``for free while collecting data for other tasks (e.g., grasping, pushing, placing). Weshow our approach achieves significantly lower root mean squared error thantraditional structured light sensors and unsupervised deep learning methods ondifficult, industry-scale jumbled bin datasets.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×