Unverified paper record
HarvestField: A Decision-Support Framework for Robot-Oriented Kiwifruit Canopy Management Using Spatial Harvestability Modelling
11 Jun 2026 · 10.21203/rs.3.rs-9528703/v1
Abstract
Abstract Purpose Robotic harvesting systems are commonly evaluated in canopies treated as fixed operating environments, although kiwifruit canopy architecture is actively shaped by seasonal pruning, fruit thinning, tipping, and cane training. This study developed HarvestField, a decision-support framework for quantifying how robot-oriented canopy management affects harvesting feasibility before field operation. Methods HarvestField defines a spatial harvestability field \(\:\mathscr{H}\left(x\right)\in\:\left[\text{0,1}\right]\) as the product of reachability, visibility, clearance, and detachability sub-fields. A mass-weighted field average at fruit positions yields the Harvested Yield Index (HYI). The framework couples an L-system-based functional–structural plant model of Hayward kiwifruit with NSGA-III multi-objective optimisation over six management variables, jointly evaluating HYI, yield, and labour cost. Results Calibration against a published 12,000-fruit robotic kiwifruit harvesting benchmark produced HYI = 0.562, with an absolute deviation of 0.004 from the observed success rate of 0.558. HarvestField-optimised management increased mean HYI relative to standard management (0.642 versus 0.587). Under equal fruit counts, \(\:\mathscr{H}\)-guided thinning improved HYI by up to 0.301 compared with uniform thinning. Sobol analysis identified clearance neighbourhood radius as the dominant parameter (\(\:{S}_{T}=0.576\)). Conclusion HarvestField provides a spatially explicit and robot-specific framework for linking kiwifruit canopy management with harvesting feasibility. It supports management trade-off analysis while indicating the need for independent orchard-scale validation before deployment.
Plant phenotyping relevance
キウイフルーツ樹冠の可視性・到達性・クリアランス・離脱性を統合して、果実位置ごとの収穫可能性を定量化する計算フレームワークが研究の中心であり、単なる収穫対象の検出を超えた植物構造由来の状態推定である。
abstractThis study developed HarvestField, a decision-support framework for quantifying how robot-oriented canopy management affects harvesting feasibility before field operation.
abstractHarvestField defines a spatial harvestability field \(\:\mathscr{H}\left(x\right)\in\:\left[\text{0,1}\right]\) as the product of reachability, visibility, clearance, and detachability sub-fields.
abstractCalibration against a published 12,000-fruit robotic kiwifruit harvesting benchmark produced HYI = 0.562, with an absolute deviation of 0.004 from the observed success rate of 0.558.
Code and data availability
植物フェノタイピング解析を再現する公開資産であることを、入力本文と直接リンクから確認できなかったため保留しました。
No evidence-backed public reproduction asset is currently recorded.
This is an automatically classified, unverified record. Curator approval is required before any resource enters the Catalog.