We will develop multiple two-arm robot modules to achieve the multiplying effect with improved harvest efficiency. The perception algorithm will be extended to enable simultaneous detection and localization of apples, branches, and foliage. An efficient multi-arm, collision-free planning algorithm will be developed to effectively coordinate the two arms for maximized harvesting efficiency. A light control system with artificial lighting will be designed to allow the robots to work under different light conditions during day and at night. Lastly, a new algorithm will be developed for integration with the current fruit detection algorithm to determine the harvest readiness of the fruits based on the color and size for selective harvesting of high-valued apple varieties. 

We will develop and evaluate a new, improved in-field pre-sorting system equipped with a new generation AI-based algorithm with the augmented capability for full inspection of apples for surface defects in addition to size and color. Images for different types of defects for several varieties of apples will be collected for building and evaluating the new AI models for defects detection. Laboratory and field tests will be conducted to determine the optimum operational parameters for the pre-sorting system. A graphical user interface (GUI) will be developed for integration with the central computer system of the AIMS. 

We will design and build a new autonomous mobile platform for integration with four 2-arm robotic harvesting modules and the new in-field pre-sorting system. The new mobile platform will be such designed that it is suitable for different orchard systems and is also environmentally friendly. New, improved bin fillers will be designed and built for handling fresh-quality and cull apples. Automated functions for handling empty and full bins will be incorporated with the mobile platform. In addition, a central power supply and computer system will be designed for coordinated, automated controls of robotic harvesting, sorting, and bin filling and handling. 

We will conduct field tests and demonstrations for the robotic harvester and AIMS in commercial orchards in Michigan, Pennsylvania, and Washington to evaluate the efficacies of the new robotic harvester and in-field pre-sorting system in high-density apple orchard systems. We will also provide in-person and virtual extension programming on harvesting technology in orchards and results from field tests of AIMS technology in commercial production systems. We will evaluate grower and industry perception and potential adoption of AIMS technology for use in the tree fruit industry. Finally, we will partner with K-12 educators and university programs to teach students about the impact of technology in agriculture and orchard systems and mentor young scientists and students in higher education. 

We will collect and label apple fruit images using various sensing systems in different geographical locations. We will develop state-of-the-art deep learning models for fruit detection, grading, counting, and localization. And finally, we will establish an open-access data sharing platform for professional communities and the public. 

We will study the economics of the AIMS under different adoption models and harvesting strategies to facilitate adoption and diffusion. We will also assess the long-term viability of the H-2A program and its potential impact on the adoption of robotic harvesting technology.