STTR Phase I: Integrating Vision-Guided Collaborative Robots for Postharvest Processing of Produce

Inversai, Inc., Athens GA

Investigators

Abstract

The broader impact of this Small Business Technology Transfer (STTR) Phase I project is to empower the processors of harvested fruits and vegetables with the flexibility to use robotic automation to meet their labor needs. The automation uses collaborative robots (cobots) guided by computer vision, which are potentially safe around humans. The technology will help assure consistent produce quality and processing rates. Through a robust cobot-based solution, the project will provide an affordable, sustainable, and safe means for farms of all sizes to keep up with their production goals, which will sustain competition and the nation’s food supply. This project has the added benefit of upskilling workers in farms by creating openings for more technically oriented positions, both in monitoring and maintaining the cobots. Instead of tediously programming the cobot for each use, the project is introducing a new way of translating the tasks performed by humans to the cobot by learning from camera recordings. It will also improve understanding of how cobots can safely be used alongside humans in a shared working space. This Small Business Technology Transfer (STTR) Phase 1 project aims to make it possible to use cobots with human workers on tasks that go beyond the traditional pick-and-place. The proposed technology will automate processing line tasks that require computer vision, which is challenging because accurate and reliable perception must guide the robot’s motion. Research has coalesced the technical challenges on the path to a viable commercial product around five steps. These start with a formal description of the task domain followed by using robust implementations of noise-tolerant machine learning algorithms for automatically learning the task, and end with a solution that integrates the learned task behavior with a vision-guided cobot system. Phase 1 will support research toward addressing two problems. The first is to design an intuitive way to elicit a precise specification of the client’s task domain. A digital conversational assistant will utilize multiple modalities for the elicitation. The second is the inability of available implementations to generate coworker-aware and efficient cobot movements. The research will investigate and develop significant improvements to the cobot motion to improve coworker safety while reducing the processing time by an expected 50%. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.