Non Technical Summary
The BerryXpert project aims to revolutionize strawberry production by developing an innovative automated runner cutting system, addressing a critical labor-intensive task in the industry. This multifaceted approach combines advanced sensing systems, robotics, and cutting-edge digital twin technology to create a comprehensive solution for efficient runner management.Our team's intellectual merit lies in the integration of diverse technologies:1. Development of a multi-modal vision system for precise runner detection and localization,2. Creation of both a mechanized disc cutter and a robotic arm for versatile cutting operations,3. Implementation of a digital twin for virtual testing and optimization, significantly reducing development time and costs.The project's broad impact extends beyond technological advancement:1. Alleviating labor shortages and reducing production costs in the $540 million Florida strawberry industry and beyond,2. Enhancing crop yield and quality through precise and timely runner control,3. Establishing a scalable framework for future agricultural robotics applications.BerryXpert's innovative approach includes a unique canopy manipulation system for runner exposure and utilizes sim-to-real transfer techniques to bridge virtual and physical environments. By validating the system across diverse strawberry varieties, we ensure adaptability to various growing conditions.This project not only addresses immediate industry needs but also paves the way for broader automation in specialty crop production. Through extensive field trials and industry engagement, BerryXpert aims to create a practical, efficient solution that will significantly enhance the sustainability and competitiveness of strawberry farming in the United States.

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
205	1122	2020	100%

Knowledge Area
205 - Plant Management Systems;

Subject Of Investigation
1122 - Strawberry;

Field Of Science
2020 - Engineering;

Goals / Objectives
This project aims to develop an innovative, advanced automated system for runner cutting in strawberry production, addressing a critical gap in current agricultural automation technology.Each objective addresses a critical aspect of the project, with interdisciplinary collaboration between experts in sensing systems, automation, and strawberry plant breeding.Objective #1: Develop Advanced Runner Detection Systems: The goal of this objective is to design and implement advanced sensing systems capable of detecting runner growth and relevant field conditions in real time with a canopy manipulation system for runner exposure. These sensors will guide the automation systems to ensure precise and timely runner cutting, reducing the need for manual labor.Objective #2: Create Automated Runner Cutting Mechanisms: This objective addresses the development of two complementary automated cutting systems: a mechanized disc cutter for non-selective cutting and a robotic arm cutter for precision cutting. These systems will be tailored to different operational needs, providing flexibility for strawberry producers.Objective #3: Develop and Implement Digital Twin Technology: This objective focuses on developing a digital twin model of the strawberry field environment. The digital twin will simulate real-world conditions, including plant growth patterns, runner positions, and field topography. This virtual environment will be used to optimize the performance of the automated systems before field trials.Objective #4: Field Validate the Integrated Automated System for Representative Strawberry Varieties: The integrated automated systems developed in Obj. 1 and 2 will be field validated to ensure effectiveness in real-world conditions, focusing on cutting accuracy, labor savings, and crop health. Strawberry growers typically use at least three different varieties in a given season, and these varieties produce different numbers of runners. They also have different canopy characteristics, optimal in-row spacings, fruit positions and other characteristics that would affect mechanical operations such as runner cutting. We will include current commercial varieties as well as promising breeding selections to represent both the present and future of Florida's strawberry industry.

Project Methods
This three-year project will develop and field-validate an automated strawberry runner cutting system by integrating advanced perception, cutting mechanisms, digital twin-enabled optimization, and multi-variety field trials. Work is organized into four objectives with iterative refinement using measurable performance criteria and stakeholder feedback.Objective 1: Develop advanced runner detection systems. We will improve plant identification and runner localization under commercial canopy conditions where plant spacing (10-12 in) and interconnected canopies reduce visibility. First, we will map plant locations early in the season before canopy closure and use these priors to maintain plant identity as canopies merge. Second, we will design a canopy manipulation module (multi-gripper, gentle "gather-and-center" motion) to expose runners on the bed surface while minimizing disturbance to fruit and foliage. Third, we will implement a multi-modal stereo vision pipeline that captures synchronized RGB and depth data to (i) detect plants and runners, (ii) quantify runners per plant, and (iii) localize runner 3D positions and classify whether runners are cuttable from bed edges or located on the bed surface. Onboard computing will enable real-time inference and decision outputs that directly inform cutting actions. Performance will be evaluated by detection accuracy (target >95% for plants and runners), runner counting accuracy (target >90%), and correct runner classification for cutting strategy.Objective 2: Create automated runner cutting mechanisms and integrate with a mobile platform. We will develop two complementary cutting solutions to address both non-selective and selective runner removal needs. (i) A mechanized disc cutter will be designed for continuous high-speed runner cutting, driven by a BLDC motor and protected by a blade safety system using distance sensing and real-time motor cutoff when approaching unsafe proximity to the bed. (ii) A robotic arm system will be integrated with the perception output to enable selective cutting, using a custom nipper-style spring-loaded end effector and stereo vision-guided 3D targeting for approach angle and cut-point selection. We will upgrade our existing autonomous mobile platform to increase payload capacity and enable GPS-guided navigation with obstacle awareness. In Year 3, we will complete system integration for both payload configurations (disc cutter + platform; robotic arm + platform) and validate closed-loop operation, including transitions between navigation and cutting modes. Key performance metrics include continuous disc cutting operation (>30 min without overheating), safety response to obstacles (≤100 ms), robotic arm positioning accuracy (±1 mm) and perception-to-action latency (<200 ms), and platform transport capability (≥1 km under simulated field conditions carrying full payload).?Objective 3: Develop and apply a digital twin to accelerate optimization and sim-to-real transfer. We will create a physics-enabled strawberry digital twin using procedural plant models (with runner parameterization) integrated into an NVIDIA Isaac Sim environment representing plasticulture beds and field layout. Digital representations of the disc cutter and robotic arm will be included to verify design constraints and to train perception and cutting strategies. We will generate synthetic datasets for runner detection and cutting point identification, refine models using augmentation and a custom loss function emphasizing runner endpoint accuracy, and implement domain randomization (lighting, plant geometry, environmental variability) to improve robustness. Finally, we will bridge simulation to the physical system through iterative deployment, performance monitoring, and model updates, tracking transfer learning efficiency and targeting real-world performance within a 20% margin of simulation across key metrics.Objective 4: Field-validate the integrated system across representative strawberry varieties. Annual field trials will be established at UF/IFAS GCREC using UF commercial varieties (e.g., 'Florida Brilliance' and Medallion™ 'FL 16.20-128') plus additional breeding selections each year to capture diversity in runnering patterns and canopy architectures. We will collect weekly runner counts in replicated sub-plots to quantify genotype- and season-dependent runner emergence and to link plant traits to operational performance. The integrated cutting systems will be deployed in these trials to evaluate autonomous operation, cutting accuracy, speed, adaptability across varieties, and labor savings compared with manual runner removal. Target field performance is ≥70% cutting accuracy during initial integration (Years 1-2) and ≥80% cutting accuracy across varieties during validation (Years 2-3), with demonstrated adaptability across at least five varieties by project end.Risk management and alternatives. Anticipated challenges include limited runner visibility under dense canopies, potential mis-assignment of neighboring runners, and the possibility of exposing leaves/fruits on the bed surface. Mitigations include improving multi-class vision to separate plant organs, using mapped plant priors to preserve identity, and refining manipulation and cutting logic to avoid non-target tissues. Weather and terrain variability will be addressed through robust sensing and operational protocols; if battery life limits field runtime, we will deploy a commercially available platform (e.g., Farm-ng) available to the team. Mechanical failures will be mitigated via a rapid maintenance protocol and redundant critical components.