Evaluation of Photoelectric Sensors for Robotic Transplanting

Abstract

The ability of a robot to intelligently transplant bedding plant seedlings was studied using photoelectric sensors to detect the seedlings and a compliant finger-type gripper. The sensors were mounted on an IBM 7535 robot (IBM, Boca Raton, FL, USA), and used reflected light in the infrared range to detect plant foliage. Three tests using two different sensor configurations were evaluated: (1) two sensors mounted side-by-side on the same side of the gripper where the sensors output was monitored once per cycle, (2) two sensors mounted side-by-side on the same side of the gripper where the sensors output was monitored twice per cycle, and (3) four sensors mounted in pairs on opposite sides of the gripper where the sensors signal was again monitored twice per cycle. Twelve different bedding plant species were transplanted to evaluate the ability of the sensors to detect the varied seedling morphologies during transplanting. Sensor outputs were used to decide upon multiple actions at two points within the robot control program: (1) just after a seedling had been retrieved from the plug flat, and (2) just after the seedling was deposited into the grow flat.