A four-arm harvesting robot system was designed to integrate with the fruit picking-collecting-transporting multifunction for the apples’ automatic harvesting. Taking the standardized tall-spindle and dwarf-rootstock apple tree as the object, the target operational area was determined for the harvesting robot, according to the fruits’ spatial distribution within the tree canopy. A new configuration of a four-arm picking manipulator and the operational mode were proposed with the four Cartesian coordinate arms in the three degree-of-freedom (DOF). An electric-pneumatic hybrid dual-stage driving structure was utilized to ensure efficient and large-scale telescopic motion within the tree canopy. Additionally, a CAN open bus-based integrated drive-control harvesting gripper was designed to enable efficient harvesting operations via a combination of fruit gripping and twisting actions. A multi-task deep convolutional network was adopted to recognize the fruit’s discrete visual pixel areas that were caused by branches and leaves occlusion. As such, the semantic segmentation of the occluded fruits and end-to-end determination of the discrete areas’ ownership were realized to overcome the traditional single-task networks in the classification of discrete regions of the same fruit. The view frustum projection model was introduced to locate the centroid of the target fruit, according to the local point cloud information on the surface. A novel strategy of four-arm picking task area partitioning was proposed, according to the clustered distribution characteristics of fruits within the tree canopy. The time-optimal four-arm collaborative picking task planning was also proposed to achieve the efficient traversal of different regions inside the tree canopy by the robotic arms. Finally, the key components of the harvesting robot were integrated to develop the autonomous harvesting workflow. The production trials were also conducted in a high-density dwarf rootstock orchard. The results showed that the recognition rate on the visible fruits was 92.94%, among which 90.27% of the fruits’ positioning accuracy was sufficient for picking operations. The robot’s average overall picking efficiency was 7.12 seconds per fruit, among which the efficiencies of single-, dual-, and four-arm were 9.59, 8.17, and 4.87 seconds per fruit, respectively. The efficiency of four-arm collaborative picking was approximately 1.96 times that of single-arm picking. The success harvesting rate of visible fruits was 82.00%, and the overall harvesting rate for all fruits inside the tree canopy was 74.56%. The success rate of harvesting reached up to 100% in the outer peripheral areas of the tree canopy where the fruits were sparse. However, the success rates of target recognition, location, and operation were significantly lower in the inner-dense region where the fruits were intensive, resulting in a harvesting success rate of 73.63%. The harvesting failures were attributed to the fruits that were obstructed by branches and leaves, leading to the visual recognition and positioning accuracy, as well as the interference and collisions with the harvesting manipulator. Therefore, the robot's capability of autonomous obstacle avoidance was enhanced to improve the tree structure and the performance of this harvesting robot. This finding can be considered as the preliminary exploration for the development and application of robotic harvesting models for freshly-eaten fruits.

Pressure atomization immunization can be confined to the high noise, uneven size distribution of droplets, and low effective inhalation rate in poultry houses. This study aims to design the piezoelectric component with the microporous ultrasonic atomization in an immunization robot. The performance of spray immunization robots was also improved in the caged poultry houses. A multi-physics coupled model of the atomizing plate was established using finite element simulation software. In the boundary conditions, the fixed constraints were applied to the edges of the atomizing plate in order to simulate the assembly clamping. While a driving voltage was applied on the upper surface of the piezoelectric ceramic, where the interface contacting the metal substrate was set as the ground potential. In the mesh generation, a physics-controlled adaptive strategy was adopted with local refinement to the key areas, such as the interface between the piezoelectric ceramic and the metal substrate. The vibration modes and frequency response were analyzed to determine the optimal resonant frequency of 113 kHz. An equivalent circuit model was used to accurately represent the piezoelectric oscillation behavior of the atomizing plate. The formula derivation showed that the equivalent impedance of the series matching network was primarily resistive when the circuit operated near the resonant frequency. At this point, there was no phase difference between the driving voltage and the input current. As such, the resonant working state was effectively determined to check whether the phase relationship between voltage and current was zero. A precision spray control was proposed using PI frequency tracking, in order to prevent some changes in the resonant characteristics of the atomizing plate due to the prolonged operation and environmental factors. The voltage and current signals were collected at both ends of the atomizing plate. The power factor angle was then calculated to compare the current phase difference with the reference value under resonant conditions. This phase error served as the input to the PI controller. A control signal was then generated after adjustment. Its output frequency was adjusted for the dynamic compensation in the attenuation of the atomizing plate's vibration. A modular component of immunization atomization was integrally designed in an spray immunization robot of a poultry house. Specifically, the spray immunization robot consisted of a mobile platform, an atomization and control system. Some sensors were also used in the crawler chassis for the autonomous movement and navigation within the poultry house. Its atomization system included a medicine tank and multiple nozzles with atomizing plates, thus forming a sealed circulation channel of medicinal fluid. A remote terminal was employed to set the parameters of the robot. Unmanned immunization spraying was then automatically carried out using the given plan. Static single-nozzle atomization tests were finally conducted to verify the immunization robot application. The results showed that the test group of PI frequency tracking achieved a 1.57% increase in the average atomization amount after the atomizing plate for 1 h, compared with the fixed resonant frequency. Some measurements were also taken with a HELOS/VARIO laser particle size analyzer at a distance of 30-50 cm from the nozzle. It was found that the proportion of droplets meeting the required size for immunization spray reached 90.81%. In the actual poultry house, the water-sensitive paper was used to capture the aggregated droplets caused by multiple airflows during the robot's movement. There was an average droplet size increase of 42.81 μm, compared with the static laboratory environment. The overall proportion of droplets with the required size for immunization spray reached 90.2%, fully meeting the droplet size requirements for poultry house spray immunization. The findings can provide the technical support for the spray immunization robots in poultry houses.

Apple harvesting is one of the most complex and least mechanized processes at present. A picking robot can greatly contribute to the advancement of the apple industry. Among them, the picking manipulator is one of the key components in the picking robot. However, the current apple picking manipulators are limited to the complex structures and low modularity, unsuitable for the multi-arm picking operations. It is necessary to develop an apple-picking manipulator with a larger range of motion, high modularity, and a lightweight structure. Efficient, stable, and flexible operation can often be required to optimize the configuration parameters of the manipulators. Particularly, the space constraints rather than motion performance have been focused primarily on optimization in recent years. In this study, a modular configuration of the apple-picking manipulator was designed to optimize the parameters of the motion performance. Firstly, an apple-picking manipulator was designed to fully meet the operational requirements, according to the distribution of the fruits in orchards. The manipulator consisted of three translational and three rotational joints. Specifically, the three translational joints were used to control the motion along the x, y, and z axes, while the three rotational joints were used to control the rotation along the roll, pitch, and yaw axes. The horizontal and telescoping joints were used to realize the different types of motion in the xoy plane using the joint drive motors 1 and 2 with the translational motion. Secondly, the multiple indices were combined with a single objective function in order to evaluate the operational accessibility, structural compactness, velocity, and load smoothness. The analytic hierarchy was employed to determine the weights of each index. The linear weighting was used to generate the objective function. As such, an optimization algorithm was then proposed using an improved hippopotamus optimization algorithm (IHOA). Among them, the hippopotamus optimization (HO) was employed for the global search in the initial stage, the particle swarm optimization (PSO) was to accelerate the convergence using collaboration and learning within the population, and the incorporated simulated annealing (SA) was to introduce the random perturbations. Finally, the simulation and field experiments were performed to validate the operational reachability, structural compactness, velocity, and load stability of the apple picking manipulator. The simulation results showed that the link lengths of the pitch joint, horizontal joint, telescoping joint, and end-effector revolute joint were 122.02, 138.00, 101.45, and 103.12 mm, respectively. The link offsets of the telescoping joint, end-effector revolute joint, end-effector prismatic joint, and twisting joint were 855.00, 166.67, 189.95, and 126.63 mm, respectively. The installation heights of the lower and the upper picking manipulator were 1344.59 and 2460.00 mm, respectively. The experimental results showed that operational accessibility index F₁, structural compactness index F₂, global velocity fluctuation performance index F₃, and global load fluctuation performance index F₄ were 97.05%, 2 882.74 mm, 0.20 m/s, and 0.15 N·m, respectively. Field experiments showed that the maximum absolute torque increments for the pitch joint, joint motor 1, and 2 with translational motion, and the end rotation joint were 0.51, 0.87, 0.80, and 0.79 N·m, respectively, during picking apples at the boundary points. The maximum absolute velocity increments were 0.03, 0.17, 0.17, and 0.01 m/s, respectively. Therefore, the manipulator demonstrated full accessibility to the boundary positions within an operational range of 890.25 to 1 035.47 mm from the tree trunk. This finding can also provide valuable insights to design the modular picking manipulators for apple harvesting.

Edible mushrooms are rich in nutrients; however, harvesting mainly relies on manual labor. Coarse localization of each mushroom is necessary to enable a robotic arm to accurately pick edible mushrooms. Previous studies used detection algorithms that did not consider mushroom pixel-level information. When these algorithms are combined with a depth map, the information is lost. Moreover, in instance segmentation algorithms, convolutional neural network (CNN)-based methods are lightweight, and the extracted features are not correlated. To guarantee real-time location detection and improve the accuracy of mushroom segmentation, this study proposed a new spatial-channel transformer network model based on Mask-CNN (SCT-Mask-RCNN). The fusion of Mask-RCNN with the self-attention mechanism extracts the global correlation outcomes of image features from the channel and spatial dimensions. Subsequently, Mask-RCNN was used to maintain a lightweight structure and extract local features using a spatial pooling pyramidal structure to achieve multiscale local feature fusion and improve detection accuracy. The results showed that the SCT-Mask-RCNN method achieved a segmentation accuracy of 0.750 on segm_Precision_mAP and detection accuracy of 0.638 on Bbox_Precision_mAP. Compared to existing methods, the proposed method improved the accuracy of the evaluation metrics Bbox_Precision_mAP and segm_Precision_mAP by over 2% and 5%, respectively.