A Robotic Solution to Finding and Removing Spotted Lanternfly Eggs
A team from Carnegie Mellon University's Robotics Institute has developed an autonomous robot to control the spread of spotted lanternflies — invasive insects known to destroy economically important crops. The team created TartanPest by mounting a robotic arm to the base of an all-electric Amiga microtractor created by CA-based robotics company Farm-ng. Watch this video to learn more about TartanPest.
"Currently, spotted lanternflies are concentrated in the eastern portion of the nation, but they are predicted to spread to the whole country," said Carolyn Alex , an undergraduate researcher on the TartanPest team. "By investing in this issue now, we will be saving higher costs in the future."
Transcript
00:00:00 foreign we have designed a project which aims to address the spotted lanternfly issue which especially impacts Farmers since they tend to go for economically important crops like Hardwoods ornamentals and grapevines this problem especially impacts our local region of Pennsylvania where it has the potential to drain over 300 million from the
00:00:30 economy annually which is why we have developed a robotic solution to this problem with the potential to lower the economic and social impacts of the spotted lanternfly on a broader scale a robotic solution to these pests can increase the efficiency of farms lower costs lower chemical pollution of crops and save labor costs for Farms the pharmacy Amiga was a fantastic platform
00:00:55 to work with during this competition it is modular and flexible so we could Mount all the equipment sensors it also allowed us to drive a road in different testing fields which was a really nice capability for a solution we've met an arm to the medical base that way the Amiga base can drive up and Target landerfly egg sacs this allows us to Target the most susceptible part of the
00:01:14 lanternfly growth cycle we've mounted a drill brush sensor onto the end of the arm which then targets the egg mass and then uses computer vision to reach up and scrape it off we trained a deep learning model particularly YOLO model we used a crowdsource data set in collaboration with Dr RT Singh from Iowa State University so they have a data set with a large number of lanternfly eggs
00:01:37 that people capture just from their cell phones and we have like 700 of those images since the data set doesn't have like millions of images that you would need for a deep Learning System we augmented that with transforms like rotation translation and zoom and things like that and we took that and trained the model and we used that model for for the detection system we used open AI
00:02:03 step API to extract the locations of the detections after identifying the Mass the next part is using that information to make the arm move to the Target switch on the brush and scrape off the egg Mass we mounted a six degree of Freedom robot arm from U Factory the exom 6 in particular we also use its control box to switch on and off the brush that we're using using a relay
00:02:31 switch simulation is a crucial component of Robotics the primary advantage of simulation is that it could streamline the development process the simulation could help us to experiment with different Hardware configuration iteratively and quickly we utilize the Gazebo simulator due to its social activity with Ross to do the simulation we need to have a new rdf file it is an
00:02:59 XML type file that will include all the physical description of a robot in this case we create a geometry by using the cad and then we took the geometry and add weight Mass joint to the robot in the URL file by having the usdf file we build from scratch we could simply change the width the lens and deployed all kinds of sensor on the robot platform currently spotted Lantern 5S
00:03:28 are concentrated in the Eastern portion of the nation but they are predicted to spread to the whole country investing into this issue now we will be saving higher costs in the future