https://pvela.gatech.edu/classes/ 2026-04-26T04:44:50+00:00 https://pvela.gatech.edu/classes/ https://pvela.gatech.edu/classes/lib/exe/fetch.php?media=wiki:dokuwiki.svg text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:01start&rev=1724204336&do=diff VisMan: Connecting and RViz ---------- Main activities are: * Log into machine and get comfortable with launching basic services. * Connect to Dynamixel servomotors, visualize joint angles and robot in rviz. * Connect to a single separate motor, send low-level commands to move it. There should be an interface for sending joint commands to the Dynamixel servomotors. Show control of one. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:02arm&rev=1724204336&do=diff VisMan: Commanding Arm Movements ---------- Now that you've gotten comfortable with linux, Dynamixel, ROS, rviz, etc., the next step is to actively move the robot arm. This involves controlling the manipulator in joint-space and in Cartesian-space. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:03cam&rev=1724204336&do=diff VisMan: Robot Eyes / Robot Vision ---------- In order to respond to environmental variations, the manipulator will need a mechanism for seeing the world and modeling what is in the scene. The easiest such sensor to work with is a depth sensor since it gives color imagery and geometric information in the form of depth. Depth data captures the structure of the world, which really means that world structure can be obtained from the depth images. Some processing is required. Let's start down the … text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:04tabletop&rev=1724204336&do=diff VisMan: Modeling the Task Workspace Plane ---------- Part of calibration also involves capturing the tabletop plane relative to the camera. It can then be used to define a collision surface in MoveIt!. Use SVD to capture plane equations. Write here the plane equations. Collect data from realsense. Perform the estimation of the frame. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:05model&rev=1724204336&do=diff VisMan: Sensing Objects in the Task Workspace ---------- Use tabletop plane height processing to isolate objects on the tabletop from the tabletop. Let them be not in contact with each other. Use OpenCV processing to get number of objects, center of objects, and any other interesting statistics. Publish a topic giving the object centroids as a point cloud or some other visualizable message type. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:06blocks&rev=1724204336&do=diff VisMan: Object Frames in the Task Workspace ---------- Use the segmented block information to recover point clouds of the objects. Estimate their reference frames using a similar SVD technique as for the tabletop plane (or maybe a PCA / eigenvector one will work). text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:07blockmove&rev=1724204336&do=diff VisMan: Perception, Planning, and Action: Pick and Place ---------- Combine everything to pick up and place blocks in the scene to a new location, possibly outside of the field of view. Demonstrate ability to pick up all blocks and place them off-screen. There is no need for them to be organized. If desired, you can put some kind of box or mini-bin there as though it were a bin storing problem. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:08blockcleanup&rev=1724204336&do=diff VisMan: Clean-Up or Clearing Task ---------- Idea is to have 2-3 different types of objects. Be able to differentiate them by size or color, then place them where they belong (different “bins” or “dump zones”). Note: We need to decide on the different types. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:09blockstack&rev=1724204336&do=diff VisMan: Stacking Task ---------- This task doesn't have dump zones, but attempts to “store” or retain some information about past place actions. Every time a piece is place in the stack, the stack model needs a height adjustment. Maintain some internal model of the stack. The simplest stack just places the piece on top of each other with same orientation. text/html 2024-08-21T01:38:56+00:00 Anonymous (anonymous@undisclosed.example.com) https://pvela.gatech.edu/classes/doku.php?id=ece4560:visman:adventures&rev=1724204336&do=diff Vision-Based Manipulation: Learning Adventures ---------- Assumptions: This series of adventures is predicated on concurrently taking the ECE4560 course, so that the material being explored maps to lecture and reading material to the degree possible. It will involve estimating reference frames for a manipulator with camera robot. These would include the arm base frame, the camera to world frame, the end-effector frame, a tabletop to camera frame, and the reference frames of objects on the tabl…