CPS: Medium: An Autonomous Robotic System for Precision and High-Throughput Tomato Phenotyping in Large-Scale Greenhouses

The goal of this project is to develop an autonomous robotic system for precision and high-throughput tomato phenotyping in large scale greenhouses. This cyber-physical system includes a mobile robot arm, a computer vision based phenotyping system, and a wireless dynamic charging system, integrating innovations spanning multiple disciplines including robotics, computer science, and power electronics. In particular, four research goals are proposed: (1) Develop deep learning models to localize the tomato plants, perceive the greenhouse environment, and select the target fruits for phenotyping by integrating domain knowledge in phenotyping and greenhouse managers’ specific needs. (2) Analyze the dexterous and safe workspace of the robot arm to locate the target fruits inside it for collecting images from different angles safely. Develop novel robot motion planning algorithms to take high quality images for phenotyping and prevent potential damage to the plants and robot. (3) Develop a multi-task learning model to compute the eleven tomato fruit traits and automatically evaluate the quality of the phenotyping results based on uncertainty analysis and domain knowledge to determine if phenotyping needs to be redone, which will close the loop of this system. (4) Develop an optimized high-efficiency, high- reliability, and low-cost wireless dynamic battery charger concept to provide power to the autonomous mobile robotic phenotyping system operating uninterruptedly in large scale and humid greenhouses to improve work efficiency and save energy costs.