Flexible control of apple picker using active disturbance rejection control and model predictive control

The apple industry has achieved large-scale production with intelligent agriculture. However, the fruit picking can still rely heavily on manual labor, particularly under the mechanized operation in the hilly and mountainous areas. Among them, the end-effector is one of the most important components to directly interact with the fruits in the apple-picking robots. The flexibility and control accuracy are often required to improve the picking efficiency for less damage to the fruits. However, the conventional end-effectors are limited to the active flexibility to different apple sizes and the resistance to complex disturbances. In this study, a technical solution was proposed to combine the structural optimization and advanced control strategy. Firstly, a lightweight and adjustable flexible end-effector was designed for the apple-picking robots. A three-jaw structure was adopted to simulate the human grasping movements. A micro servo electric cylinder was used as the driving mechanism. The translational motion of the connecting rod was then converted into the opening and closing motion of the flexible gripper. A four-speed adjustable gear structure on the passive finger base was utilized to realize the rapid gear switching within 15 s after a tenon-mortise slot connection. The end-effector safely grasped the apples with diameters ranging from 20 to 110mm. The components were fabricated by 3D printing technology. The gripper fingers were made of silica gel-fiber composite (soft material) with the fin effect, effectively protecting the apple peel from damage. Secondly, an active disturbance rejection control (ADRC)-model predictive control (MPC) force-position hybrid control architecture was integrated with Position-Velocity Calculation (CPV) and Linear State Error Feedback (LSEF), in order to enhance the control accuracy under the friction disturbances during transmission. An Extended State Observer (ESO) of the ADRC was used to real-time estimate and compensate for the total disturbances, including the friction, model errors, and external interference. The CPV module also replaced the conventional Tracking Differentiator (TD) to avoid the phase lag and noise sensitivity. An accurate mapping relationship among force, displacement, and velocity was established using kinematic equations and Newton’s second law; The LSEF module served as the coordination hub between MPC and ESO. Prediction deviations were dynamically corrected using speed error terms. The measured forces were introduced to enhance the disturbance suppression; The MPC module was used to optimize the control strategy using a rolling time-domain mechanism. The disturbance feedforward compensation items were embedded in the prediction model. The optimal control was achieved under constraints. Finally, the simulation and physical experiments were conducted to verify the effectiveness of the scheme. The results show that the derived forward and inverse kinematic models of the end-effector were highly consistent with the simulation, providing a reliable theoretical basis for the practical control. Physical experiments were carried out under four gradient target grasping forces (2.5, 3.0, 3.5, and 4.0 N) using standard apples with a diameter of 55.59 mm. The experimental results indicate that the system shared an excellent dynamic response and control accuracy. Specifically, the average response time of the force tracking was 0.85 s, the average steady-state time of the contact force was 3.08 s, and the error of the dynamic force control was stably maintained within ±0.1N. Ablation experiments confirm that the MPC module significantly shortened the force tracking response and convergence time, while the LSEF module was designed for engineering practicality and parameter robustness with a simple linear structure. Comparative experiments with ADRC+PSO and DLADRC-OBLHOA control schemes show that the ADRC-MPC architecture outperformed the rest schemes, in terms of both steady-state arrival time and force tracking response time, with a reduction of 35.3% and 46.2%, respectively, compared with the ADRC+PSO scheme. The conventional end-effectors with fixed sizes were overcome for the better adaptability of the control strategy to complex disturbances during picking. A feasible and efficient technical solution was achieved for the flexible and precise picking of the apple-harvesting robots. This finding can also provide a strong reference for the similar harvesting of the ellipsoidal fruits, such as pears and citrus.