Design and experiments of an automatic seed potato release mechanism for plastic film mulching cultivation

Abstract

Manual management of seedling emergence in plastic-mulched potato fields is inefficient and labor-intensive. To address this, we developed a four-degree-of-freedom (4-DOF) robotic manipulator for automated seedling release and an improved YOLOv5s-se model for detecting seedlings beneath plastic mulch in complex scenes. A serial 4-DOF kinematic model was established using Denavit–Hartenberg (D–H) parameters, validated via forward and inverse kinematic solutions, and its workspace was analyzed with the MATLAB Robotics Toolbox. Simulation results show that cubic and quintic polynomial interpolations yield path-consistent trajectories; however, analysis of joint angles, angular velocities, and angular accelerations indicates that the quintic interpolation provides superior motion and overall trajectory smoothness, with joint velocities < 2.3 rad·s −1 and accelerations < 7.2 rad·s −2 . Linking the simulation and field evaluation, we treat the field seedling-damage rate as the operational error indicator; it measured 1.33 % in our tests. In field trials, the improved YOLOv5s-se achieved a 94.67 % recognition rate for under-mulch seedlings; the robotic system reached a 90.67 % seedling-release success rate with an average throughput of 12.32 plants·min⁻¹; and the mechanical structure met the design requirements for strength, stiffness, and workspace. This study provides a theoretical foundation and technical support for developing automation equipment for plastic-mulched potato cultivation.