Robotic Manipulation and Physical Interaction in Agricultural Field Environments

Agricultural robotics is advancing rapidly as growers face labor shortages, rising production costs, climate variability, and pressure toward more sustainable and precise practices. Navigation and sensing have received sustained attention, and platforms that traverse and map field environments are now widely reported. Physical interaction with the crop has advanced more slowly. A manipulator must locate a target partly hidden by leaves, estimate its pose and compliance, reach it through a canopy that pushes back, detach it without damaging fruit or plant, and do so at a rate that makes economic sense. These difficulties compound over a season: a system tuned to one cultivar at one growth stage often degrades as the canopy fills in, as fruit load changes, or when the same task is attempted at another farm. Comparison across groups is difficult as well, since evaluation protocols and success criteria vary widely and supporting data are rarely released. How manipulation can be made robust to biological and seasonal variation, transferred between crops and farms with limited new data, and evaluated in ways that support genuine comparison remain open questions.This Research Topic focuses on robotic manipulation and physical interaction with crops in real agricultural environments — harvesting, pruning, thinning, weeding, targeted spraying, and crop handling, where a robot must make deliberate contact with material that is deformable, easily damaged, occluded, and changing through the season. Perception, localization, navigation, motion planning, and control fall within scope where they support such manipulation, particularly where the work examines how performance holds across growth stages, cultivars, seasons, and fields. Submissions reporting new systems or methods should be validated on real crops in an operational field, orchard, greenhouse, or comparable setting, and should release the supporting datasets, code, or both. We ask that results be reported together with the operating conditions under which they were obtained, using whichever measures suit the task; success rates, cycle times, throughput, crop damage, robustness to disturbance, and failure modes are common examples.This Research Topic welcomes contributions addressing, but not limited to:• Manipulator design, end-effectors, and grasping of deformable and easily damaged crops•Harvesting, pruning, thinning, weeding, targeted spraying, and post-harvest crop handling•Compliant, force-aware, and contact-rich control for interaction with plants and terrain, including adaptive control under uncertainty•Manipulation-oriented perception: detection, pose estimation, and occlusion reasoning that transfers across cultivars and growth stages•Localization, mapping, and state estimation that stay valid as canopy and field layout change through the season•Motion planning, navigation, and whole-body coordination for mobile manipulators operating among plants•Learning-based manipulation policies, sim-to-real transfer, and adaptation to new crops or fields from limited data•Multi-robot coordination and human-robot collaboration where crop manipulation is the primary task•Cross-season and cross-field benchmarking: protocols, metrics, failure and damage reporting, open datasets, and codeWe welcome Original Research, Reviews, Mini Reviews, Methods, Perspectives, Hypothesis and Theory, and Technology and Code articles. Topic editors: Mingfeng Wang, Chen Peng, Peng Wei Source: https://www.frontiersin.org/research-topics/85663/robotic-manipulation-and-physical-interaction-in-agricultural-field-environments