Poster for our presentation on our Autonomous Robotic Arm that detects objects with an overhead camera, uses motion planning and reverse kinematics to retrieve objects, and places the objects in a bin. This project was done as part of the EECS 498 Autonomous Robotics Lab at the University of Michigan.
1. Autonomous Object Retrieval with a Robotic Arm!
!
Samantha Luber, Zach Oligschlaeger, John Scheible!
Abstract! Block Detection!
Through incorporating various • Uses a Union Find algorithm
object detection algorithms on to build up pixel blobs of the
images from an overhead same color!
camera, we are able to efficiently • Determines blobs within a
track balls and blocks in the color and size threshold to be
robotic arm’s search space. By blocks!
applying a planning algorithm, • Calculates block orientation
the robotic arm is able to retrieve with respect to the arm!
reachable objects and move
them to the collection bin. !
Ball Detection! Planning!
• Optimized template matching • Detects reachable objects!
algorithm scans through • Calculates movement
images from overhead camera! required by each joint needed
• Searches for balls using a to reach object and plans joint
stored yellow ball image as the movement to avoid collisions!
template! • Detects if the object was
missed and re-attempts
Arm Communication! retrieval if necessary!
• Moves object to static goal
• LCM message passing by subscribing and
location!
posting movement commands to each
servo’s channel!