Five Questions for Yu Jiang
Yu Jiang joined the Horticulture Section at Cornell AgriTech as an Assistant Research Professor of Systems Engineering and Data Analytics for Specialty Crops in December 2019. He has a primary research appointment to develop and deploy the latest sensing, automation, and computational technologies for specialty crop research and production. He leads the Cyber-Agricultural Intelligence and Robotics (CAIR) Laboratory that combines advanced engineering and plant science knowledge to deliver optimal digital agriculture solutions for New York stakeholders and beyond. Prior to joining Cornell, Yu earned his Ph.D. degree in Agricultural and Biological Engineering from the University of Georgia in 2019. Before that, he completed his undergraduate and master studies in China, majoring in Computer Science. He is excited and thrilled to work with many great grape researchers and extension educators at AgriTech and contribute to the grape community in New York and worldwide!
What inspired you to get involved with robotics, sensors and data analytics in specialty crops?
I was born and raised in urban areas in China and my way towards agriculture in general can be traced back to my master’s study where for the first time, I witnessed the challenges in agriculture and realized the importance of using technologies to overcome them. I first learned the term “specialty crop” when I moved to Georgia. I spent half of my Ph.D. time on the development of sensing and automation techniques for improving blueberry mechanical harvestability, harvesting efficiency, and postharvest quality in the US. In fact, it was shocking to me that specialty crops with high economic and nutritional added values do not gain attention and investment as much as commodity crops - even in the US, one of the world-leading countries in agricultural technology. This makes me rethink the global food security issue from different perspectives (small to mid-sized growers and nutritional security) and motivates me to focus on specialty crops.
You were hired to apply robotics, sensor data, and data analytics approaches to several fruit and vegetable crops at Cornell AgriTech. What is your vision for the use of these tools in specialty crop research and production?
The overarching goal of my program is to integrate sensing, data analytics, and robotics technologies for new engineering solutions that can make the research and production for specialty crops more data-driven, intelligent, and systematic in the next decade.
From the sensing and data analytics perspective, I anticipate developing information systems consisting of sensing networks, AI-based models, and enhanced reality-based user interface, so data can be acquired, analyzed, and shared through all important stages from pre-season evaluation, in-season growth monitoring, harvesting, and postharvest assessment. The system will help researchers to accelerate research cycles and knowledge discovery and dissemination, so growers and stakeholders can adopt the best practices and/or intelligent decision making to solve challenging issues from the ever-changing environment.
From the manipulation perspective, robotic techniques, or automation in general, will be further specialized for crucial tasks in the special crop industries such as harvesting and pruning. These would help growers and stakeholders to deal with a foreseeable reduction of labor availability for agriculture. Combining the information systems and robotic actuators will assist the specialty crop industries as a whole to address both environmental and socio-economic challenges.
What projects are going on in your program right now?
Everyone might recognize the value of having agricultural robots and digital agriculture tools during this unprecedented global pandemic more than any time in the past. Currently, I am conducting several projects that focus on the development of robots and data analytical tools for apples and grapes. Based on the techniques used, the projects can be categorized in thre general areas: sensing, data analysis, and robotics.
For the sensing part, I am developing a 3D-imaging approach that can identify, geo-reference, and characterize apple trees in the field. This is a part of technical innovations in a big project led by Dr. Terence Robinson for precision crop load management for apples. We are also extending the 3D system for some weed science projects in collaboration with Dr. Lynn Sosnoskie.
For the data analysis part, my lab has been exploring various deep neural networks, a promising AI technique, for analyzing images to extract crop information for decision making. So far, we have some promising results and keep pushing the performance boundary for practical applications.
For the robotics part, we have received funding through the Cornell AgriTech Director’s Venture Fund to develop a ground autonomous robot that can perform versatile tasks for vineyard management. We just completed and successfully tested the robot’s autonomous navigation function in several research fields at Cornell AgriTech. This robot platform will be configured with various sensors and actuators this Fall for addressing some critical issues encountered by New York grape growers.
How will your research benefit the grape industry?
I anticipate that new engineering tools from my program will be useful for many grape research programs at AgriTech to find new knowledge beneficial to various aspects of the grape industry. Meanwhile, I look forward to receiving feedback from collaborating researchers, extension educators, growers, and stakeholders to improve designs and deliver prototypes that are effective, efficent, and affordable to fulfill the grape industry needs.
A particular effort, after I joined Cornell AgriTech, is to develop a neural network that can identify and quantify grape downy mildew in images collected using a stereo camera. Our network has been tested on the 2019 dataset and achieved about 90% accuracy. We are collecting more data this season to validate the achieved performance. In addition, we are working on a special network design to make it work in near real time, so the disease detection and quantification can be performed in the field. Combining the trained neural network and the autonomous robot, we plan to develop the first prototype of a scouting robot for grape downy mildew. This would help my collaborators (e.g., Dr. Kaitlin Gold) to speed up studies and experiments related to disease resistance and management practices locally as well as open the possibility of linking the ground proximal sensing with satellite-based remote sensing for grape disease management at larger scales.
What project would you pursue, if funding was unlimited?
Unlimited funding is particularly a luxury for an agricultural engineer, who usually needs to consider a lot of the cost because affordability can be a deal breaker for an agricultural technique. If I were lucky enough to have unlimited funds, I would integrate all fancy technologies (e.g., 5G and WiFi-6 communication, human-like robots, full-spectrum optical sensors, powerful edge-computing devices, etc.) to explore and demonstrate what we could ultimately achieve for the future agricultural production practices. This demonstration system can help us to reshape our roadmap of technical development. More importantly, it can serve as a model system to attract, encourage, and inspire the young generations to study and work on agriculture-related disciplines and/or technologies that can solve agricultural challenges.