A vision-based robotic system for precision pollination of apples

TL;DR

This study develops an integrated ground-based robotic pollination system for apples, combining a machine-vision pipeline, a 6-DOF manipulator, and an electrostatic pollen sprayer to enable precision cross-pollination. In Honeycrisp, a 2 g/L pollen suspension yielded a fruit set of 34.8% on sprayed flowers and 87.5% of clusters, while natural pollination achieved 43.1% and 94.9% respectively; in Fuji, robotic pollination underperformed relative to natural pollination with 7.2% flowers and 20.6% clusters setting fruit versus 33.1% and 80.6% for natural pollination, though fruit quality was generally comparable. The system achieved a mean machine-vision AP of 0.89 and a cycle time of 6.5 s per cluster, illustrating promising potential for scalable robotic pollination with notable cultivar- and environment-specific limitations. The work highlights the feasibility of replacing or supplementing natural pollinators in orchards, while identifying key improvements in pollen delivery efficiency, speed, and multi-variety validation required for commercial deployment.

Abstract

Global food production depends upon successful pollination, a process that relies on natural and managed pollinators. However, natural pollinators are declining due to factors such as climate change, habitat loss, and pesticide use. This paper presents an integrated robotic system for precision pollination in apples. The system consisted of a machine vision system to identify target flower clusters and estimate their positions and orientations, and a manipulator motion planning and actuation system to guide the sprayer to apply charged pollen suspension to the target flower clusters. The system was tested in the lab, followed by field evaluation in Honeycrisp and Fuji orchards. In the Honeycrisp variety, the robotic pollination system achieved a fruit set of 34.8% of sprayed flowers with 87.5% of flower clusters having at least one fruit when a 2 gm/l pollen suspension was used. In comparison, the natural pollination technique achieved a fruit set of 43.1% with 94.9% of clusters with at least one fruit. In Fuji apples, the robotic system achieved lower pollination success, with 7.2% of sprayed flowers setting fruit and 20.6% of clusters having at least one fruit, compared to 33.1% and 80.6%, respectively, with natural pollination. Fruit quality analysis showed that robotically pollinated fruits were comparable to naturally pollinated fruits in terms of color, weight, diameter, firmness, soluble solids, and starch content. Additionally, the system cycle time was 6.5 seconds per cluster. The results showed a promise for robotic pollination in apple orchards. However, further research and development is needed to improve the system and assess its suitability across diverse orchard environments and apple cultivars.

Figures (12)

• Figure 1: Overall flowchart of the proposed robotic pollination system. The system included a machine vision system with a RGB-D camera and image processing pipeline, and a mechatronic system with robotic manipulator, motion planning algorithms, and an electrostatic sprayer-based end-effector system.
• Figure 2: Robotic pollination system setup during the field trial in Naches and Pullman, WA. The processing laptop, machine vision, manipulation, and end-effector system were placed on a utility cart with an accompanying air compressor that supplied pressurized air for air-assisted atomization of pollen suspension. The electrostatic sprayer nozzle and the Intel RealSense D435i RGBD camera were rigidly attached to the distal end of the UR5e manipulator.
• Figure 3: Hardware components of the proposed robotic pollination system. The components were categorized into machine vision system, manipulation system, end-effector system, computation system, and power supply. Dell Alienware 15R4 laptop was utilized as a central control unit that handled processing, communication, and high-level control of the vision, manipulation, and end-effector systems.
• Figure 4: Electrostatic sprayer system used in this study. The pollen suspension was atomized through an air-assisted atomization process before being positively charged by the high-voltage electrode at the nozzle. The pump power supply was controlled to achieve the desired pollen flow rate.
• Figure 5: Software framework of the proposed robotic pollination system structured into three core components: machine vision, planning, and actuation. The machine vision system was responsible for tasks such as flower cluster identification, delineation, and pose estimation. A dedicated "Pollination State Machine" was developed within the planning framework to gather relevant information from the machine vision system and facilitate the planning and navigation of the robotic manipulator and actuation of the end-effector.