Design and experiment of the dedicated robotic arm for picking-and-placing of pepper plug seedlings in transplanters

Transplanters have been widely used to complete seedling picking and planting using Cartesian coordinate, serial and parallel configuration robotic arms. However, their redundant degrees of freedom and heavy weight cannot fully meet the lightweight and high flexibility requirements, particularly for seedling picking and planting of small vegetables. In this study, a lightweight and highly integrated robotic arm was proposed for seedling picking and planting. The 2ZS-Q2 semi-automatic vegetable transplanter platform was selected to design a small robotic arm with a P-R-R-R-R serial configuration (one translation and four rotations) and a flexible pneumatic seedling clamping. A systematic analysis was conducted to balance the seedling picking and planting operation space, according to the path requirements of the robotic arm. A dual-layer control architecture with Orange Pi and ESP32 was constructed to realize the real-time trajectory resolution and multi-axis coordinated motion of the robotic arm. Among them, an Orange Pi was taken as the upper-level computational unit, while an ESP32 was taken as the motion control core. The total weight of the robotic arm was 14 kg. The forward kinematic model was established using the improved Denavit-Hartenberg (D-H). The rational workspace was verified after motion simulation. The analytical solution of inverse kinematics was derived for a motion control system. Furthermore, a three-directional ranging and positioning was evaluated at the end-effector of the robotic arm, according to the three-dimensional orthogonal ranging. VL53L0X miniature laser sensors were integrated on three orthogonal planes, thus transmitting pose data wirelessly. A field test was also conducted to evaluate positioning and trajectory tracking accuracy. Results showed that the average absolute positioning error was 3.5 mm, and the global average positioning error was 3.6 mm across the seedling tray. The average tracking errors were 3.6 and 5.0 mm, respectively, for linear and curved trajectories. All accuracy indicators met the practical requirements for picking-and-placing operations. Parameter optimization experiments were conducted on the robotic arm during picking and placing. Taking "Shachen No.4" pepper plug seedlings as the test subject, a series of tests was also conducted using the response surface method, with picking frequency, path time ratio, and control frequency as factors, while the picking-and-placing success rate was the indicator. It was found that both picking frequency and path time ratio shared a significant interaction on the success rate. Once the picking frequency was too high or when the path time ratio deviated from the optimal range, the success rate decreased significantly. Excessively low control frequency and path time ratio also reduced the success rate, leading to robotic arm stuttering and seedling-gripping claw vibration. A combination of optimal parameters was determined as follows: picking frequency of 30 plants/min, path time ratio of 30%, and control frequency of 25 ms. The average picking-planting success rate was 96.0%, with a standard deviation of 1.2% and the coefficient of variation of 1.3%, which was in good agreement with the predicted values. These findings can provide potential insights for applying the picking-and-placing equipment.