Challenges and Solutions in Designing an End-effector for Robotic Cotton Harvesting

Abstract

<b><sc>Abstract.</sc></b> <b>Most U.S. cotton is harvested with very large mechanical harvesters, which are very heavy and expensive. These machines can compact the soil, leading to a decrease in water and fertilizer use efficiencies. Also, cotton plants must be defoliated and harvested at one time, so many opened cotton bolls have to wait for extended periods until harvest, resulting in reduced lint quality. If small robots could be used for cotton harvesting, these issues could potentially be addressed. The robots could potentially go to the field multiple times and harvest the cotton bolls as soon as they open. Also, such small robots would be relatively light and thus not contribute to soil compaction. </b> <b>A key component of a potential robotic cotton harvester is its end-effector. Multiple end-effector concepts were proposed and evaluated, and an optimal idea was selected for further development. Several prototypes and configurations of the selected concept were constructed and evaluated. The final prototype included 3-D printed fingers. Needles are attached to a timing belt and rotate around the finger. Increasing the number of fingers provided a superior multi-finger end-effector. A robotic test platform consisting of three linear actuators and a Jetson TX2 computer was designed to perform more tests on the end-effector. The end-effector was attached on the Z-axis. Testing showed that the end-effector could pick the seed cotton, but if it could move in three dimensions and rotate during operation, it could likely pick all the seed cotton from a cotton boll. </b>