1 of 1

Autonomous Robot Control for Spotted �Lanternflies

​

TartanPest

​

TartanPest offers a novel, cost-effective and scalable approach to addressing the challenge of invasive species through direct mechanical decimation. Its potential benefits encompass environmental conservation, economic advantages for farmers,

Acknowledgements

The proposed solution will consist of an Amiga with an attached manipulator arm.

​

​

​

​

​

The Amiga will perform the following tasks:

  1. Navigate in a specified area to search for SLF egg masses.
  2. Utilize a stereo camera to localize the position of each egg mass in 3D relative to the robot.
  3. Track the detected eggs along with their 3D position.
  4. Perform alignment maneuvers on the robot to ensure the eggs are within reach of the manipulator’s arm.
  5. Plan and execute a path for the manipulator arm to position the end effector near the eggs.
  6. Use a wire brush end effector that rotates to scrape off the egg masses from hard surfaces.

​

​

Evaluation Aspects and Results

Detection: The accuracy of spotted lanternflies' egg masses

​

​

​

Problem Statement and Motivation

Conclusion

Design Plan

The fact that spotted lanternflies could drain the state’s economy of at least $324 million annually and cause the loss of about 2,800 jobs and killing one spotted lanternfly prevents 30 to 60 insects from hatching next year motivated us to address this issue. We aimed to utilize our robotics knowledge to find a solution for this major pain point in our region.

Spotted lanternflies (SLF)  are invasive insect species that pose a significant agronomic and business problem for farmers, as their infestation can result in reduced crop yield and economic losses.

To augment the dataset, we applied techniques such as rotation, scaling, skewing, and illumination changes, resulting in an expanded dataset of around 3000 images. We evaluated the performance of our SLF egg mass detector, utilizing the YoloV5s model, based on three key metrics: mean average precision (mAP), precision, and recall. The final model achieves 66.3% in mAP, 74% in precision, and 63.4% in recall.

​

​

​

Removal: The accuracy of removing lanternflies’ egg area

​

​

We evaluate the removal accuracy by visual inspection since the

egg would be destroyed even if there is some residue, and we concluded that our scraping mechanism was always effective in removing the egg mass.

​

and the streamlining of human expertise in pest control. By deploying this robotic solution, we can foster sustainable practices in agriculture while effectively safeguarding ecosystems from the detrimental impacts of invasive species.

AI Institute for Resilient Agriculture (AIIRA) is supported by the National Science Foundation (NSF) and United States Department of Agriculture - National Institute of Food and Agriculture (award #2021-67021-35329)