Development of a Robotic Platform for Pollination of Garden Strawberry in City-Farm Conditions Using Neural Network Object Detection

Abstract

Effective pollination represents a critical biotic constraint in the cultivation of garden strawberry (Fragaria × ananassa Duch.) in vertical city farms. Traditional pollination methods demonstrate low efficiency or are impractical under multi-tier protected cultivation structures. This study presents the concept of a robotic platform equipped with an integrated computer vision system based on a neural network for autonomous detection of strawberry inflorescences and their mechanical pollination. A specialized image dataset of strawberry inflorescences was compiled to train the detection model, ensuring reliable target recognition under real greenhouse lighting conditions characterized by a dual-peak red–blue spectrum and dense plant arrangement. A manipulator prototype was developed, incorporating a 3D-printed frame, stepper motors, and a control board with an embedded neural processing unit (NPU). The proposed design is expected to provide precise positioning of the end effector and gentle interaction with inflorescence. The modular architecture of the platform enables adaptation to other crops and additional agricultural tasks. The developed solution establishes a foundation for the implementation of automated pollination technologies in protected cultivation within the framework of digital transformation of the agro-industrial sector.