This paper addressed a well-documented open problem regarding multiple unmanned aerial vehicles (multi-UAVs) formation keeping. A cooperative guidance control method based on the back-stepping ...approach is proposed to rapidly form the desired formation and reach the multi-UAV steady state. The UAV formation system consists of four UAVs, forming a regular triangle formation. One of the UAVs is a virtual leader and is located at the center of the triangle; the other three UAVs are located at the vertices of the triangle. The forward speed of leader is used as the forward direction of the formation, and the followers follow the leader in a formation flight. The error dynamics model of each follower is established by using the leader guidance mechanism, and the communication mode between any two UAVs is established based on graph theory. In addition, the guidance control law is obtained via the back-stepping approach. We reasonably construct the Lyapunov function to prove the stability of the proposed cooperative guidance control law in formation aggregation and keeping. Moreover, the proposed method is compared with the model prediction method (MPC) and the Lyapunov method to further verify the effectiveness of the proposed method. The simulation results show that each UAV can converge to the desired motion trajectory and fly in the desired formation with fast convergence speed and small steady-state error.
•A module was constructed using NIR cameras and a diffuse illumination chamber.•A YOLO V4 network was proposed for on-line apple defects detection.•The network was compacted and simplified by channel ...and layer pruning methods.•The proposed method was more promising than other object detection methods.•The proposed method was not affected by the variation of apple skin color.
In order to realize on-line detection of defective apples on a two-lane fruit sorting machine, an inspection module was constructed using NIR cameras and a diffuse illumination chamber. A real-time apple defects inspection method was proposed based on YOLO V4 deep learning algorithm. The input images were generated by combining NIR images in three consecutive rubber roller stations. Channel pruning and layer pruning methods were used to simplify the YOLO V4 network and accelerate the detection speed. A non-maximum suppression (NMS) method based on L1 norm is proposed to remove redundant prediction box after fine-tuning the pruned network. The test results indicated that the model size and inference time of the pruning-based YOLO V4 network was decreased by 241.24 megabyte (MB) and 10.82 ms, respectively, and the mean average precision (mAP) was increased from 91.82% to 93.74%, compared with the YOLO V4 network before pruning. The pruning-based YOLO V4 network based on NIR images was not affected by the variation of skin color and suitable for detects identification of different cultivars including ‘Fuji’ apple covered in red-yellow striping and red blush, ‘Golden Delicious’, and ‘Granny Smith’, with the average detection accuracy of 93.9% at the on-line test assessing five fruit per second. The overall results showed that the proposed pruning-based YOLO V4 network combined with the developed inspection module, had great potential to be implemented in commercial fruit packing line for fruit defects identification.
This paper addresses a local minima problem for multiple unmanned aerial vehicles (UAVs) in the process of collision avoidance by using the artificial potential field method, thereby enabling UAVs to ...avoid the obstacle effectively in 3-D space. The main contribution is to propose a collision avoidance control algorithm based on the virtual structure and the "leader-follower" control strategy in 3-D space that can avoid the obstacle effectively and then track the motion target. The three UAVs constitute the regular triangular formation as the control object, the virtual leader flight trajectory as the expected path, the obstacles as the simplified cylinders, and the artificial potential fields around them as approximately spherical surfaces. The attractive force of the artificial potential field can guide the virtual leader to track the target. At the same time, the follower tracks the leader to maintain the formation flight. The effect of the repulsive force can avoid the collision between the UAVs and arrange the followers such that they are evenly distributed on the spherical surface. Moreover, the follower's specific order and position are not required. The collision path of the UAV formation depends on the artificial potential field with the two composite vectors, and every UAV may choose the optimal path to avoid the obstacle and reconfigure the regular triangular formation flight after passing the obstacle. The effectiveness of the proposed collision avoidance control algorithm is fully proved by simulation tests. Meanwhile, we also provide a new concept for multi-UAV formation avoidance of an obstacle.
This paper addresses a collision avoidance problem for multiple unmanned aerial vehicles (UAVs) in the process of high-speed flight, thereby enabling UAV cooperative formation flight and effective ...mission completion. The main contribution is to propose a collision avoidance control algorithm for a multi-UAV system based on a bi-directional network connection structure. To effectively avoid collisions between UAVs and between UAVs and obstacles, the proposed consensus-based algorithm and a "leader-follower" control strategy are simultaneously applied for UAV formation control to ensure the convergence of the formation. Each of the UAVs has the same forward velocity and heading angle in the horizontal plane, and they maintain a constant relative distance in the vertical direction. This paper proposes a consensus-based collision avoidance algorithm for multiple UAVs based on an improved artificial potential field method. Simulation tests involving multiple UAVs were performed to validate the proposed control algorithm and to provide a reference for engineering applications.
► Walking FSSW produces a little higher joint strength compared with normal FSSW. ► The dwell time more than 5
s has no effects on joint strength. ► The lower sheet material underneath the hook ...hasn’t flow into the upper sheet.
Friction stir spot welding (FSSW) is a newly-developed solid state joining technology. In this study, two types of FSSW, normal FSSW and walking FSSW, are applied to join the 5052-H112 aluminum alloy sheets with 1
mm thickness and then the effect of the rotational speed and dwell time on microstructure and mechanical properties is discussed. The lower sheet material underneath the hook didn’t flow into the upper sheet due to the concave surface in the shoulder and groove in the anvil. The hardness profile of the welds exhibited a W-shaped appearance and the minimum hardness was measured in the HAZ. The results of tensile/shear tests and cross-tension tests indicate that the joint strength decreases with increasing rotational speed, while it’s not affected significantly by dwell time. At the rotational speed of 1541
rpm, the tensile/shear strength and cross-tension strength reached the maximum of 2847.7
N and 902.1
N corresponding to the dwell time of 5
s and 15
s. Two different fracture modes were observed under both tensile/shear and cross-tension loadings: shear fracture and tensile/shear mixed fracture under tensile/shear loadings, and nugget debonding and pull-out under cross-tension loadings. The performance of the welds plays a predominant role in determining the type of fracture modes. In addition, the adoption of walking FSSW brings unremarkable improvements in weld strength.
Atypical porcine reproductive and respiratory syndrome (PRRS), which is caused by the Chinese highly pathogenic PRRS virus (HP-PRRSV), has resulted in large economic loss to the swine industry since ...its outbreak in 2006. However, to date, the region(s) within the viral genome that are related to the fatal virulence of HP-PRRSV remain unknown. In the present study, we generated a series of full-length infectious cDNA clones with swapped coding regions between the highly pathogenic RvJXwn and low pathogenic RvHB-1/3.9. Next, the in vitro and in vivo replication and pathogenicity for piglets of the rescued chimeric viruses were systematically analyzed and compared with their backbone viruses. First, we swapped the regions including the 5'UTR+ORF1a, ORF1b, and structural proteins (SPs)-coding region between the two viruses and demonstrated that the nonstructural protein-coding region, ORF1b, is directly related to the fatal virulence and increased replication efficiency of HP-PRRSV both in vitro and in vivo. Furthermore, we substituted the nonstructural protein (Nsp) 9-, Nsp10-, Nsp11- and Nsp12-coding regions separately; or Nsp9- and Nsp10-coding regions together; or Nsp9-, Nsp10- and Nsp11-coding regions simultaneously between the two viruses. Our results indicated that the HP-PRRSV Nsp9- and Nsp10-coding regions together are closely related to the replication efficiency in vitro and in vivo and are related to the increased pathogenicity and fatal virulence for piglets. Our findings suggest that Nsp9 and Nsp10 together contribute to the fatal virulence of HP-PRRSV emerging in China, helping to elucidate the pathogenesis of this virus.
Abstract Multimodal deep learning plays a pivotal role in supporting the processing and learning of diverse data types within the realm of artificial intelligence generated content (AIGC). However, ...most photonic neuromorphic processors for deep learning can only handle a single data modality (either vision or audio) due to the lack of abundant parameter training in optical domain. Here, we propose and demonstrate a trainable diffractive optical neural network (TDONN) chip based on on-chip diffractive optics with massive tunable elements to address these constraints. The TDONN chip includes one input layer, five hidden layers, and one output layer, and only one forward propagation is required to obtain the inference results without frequent optical-electrical conversion. The customized stochastic gradient descent algorithm and the drop-out mechanism are developed for photonic neurons to realize in situ training and fast convergence in the optical domain. The TDONN chip achieves a potential throughput of 217.6 tera-operations per second (TOPS) with high computing density (447.7 TOPS/mm 2 ), high system-level energy efficiency (7.28 TOPS/W), and low optical latency (30.2 ps). The TDONN chip has successfully implemented four-class classification in different modalities (vision, audio, and touch) and achieve 85.7% accuracy on multimodal test sets. Our work opens up a new avenue for multimodal deep learning with integrated photonic processors, providing a potential solution for low-power AI large models using photonic technology.
Accurate estimation of forest carbon storage is essential for understanding the dynamics of forest resources and optimizing decisions for forest resource management. In order to explore the changes ...in the carbon storage of Pinus densata in Shangri-La and the influence of topography on carbon storage, two dynamic models were developed based on the National Forest Inventory (NFI) and Landsat TM/OLI images with a 5-year interval change and annual average change. The three modelling methods used were partial least squares (PLSR), random forest (RF) and gradient boosting regression tree (GBRT). Various spectral and texture features of the images were calculated and filtered before modelling. The terrain niche index (TNI), which is able to reflect the combined effect of elevation and slope, was added to the dynamic model, the optimal model was selected to estimate the carbon storage, and the topographic conditions in areas of change in carbon storage were analyzed. The results showed that: (1) The dynamic model based on 5-year interval change data performs better than the dynamic model with annual average change data, and the RF model has a higher accuracy compared to the PLSR and GBRT models. (2) The addition of TNI improved the accuracy, in which R2 is improved by up to 10.48% at most, RMSE is reduced by up to 7.32% at most, and MAE is reduced by up to 8.89% at most, and the RF model based on the 5-year interval change data has the highest accuracy after adding TNI, with an R2 of 0.87, an RMSE of 3.82 t-C·ha−1, and a MAE of 1.78 t-C·ha−1. (3) The direct estimation results of the dynamic model showed that the carbon storage of Pinus densata in Shangri-La decreased in 1987–1992 and 1997–2002, and increased in 1992–1997, 2002–2007, 2007–2012, and 2012–2017. (4) The trend of increasing or decreasing carbon storage in each period is not exactly the same on the TNI gradient, according to the dominant distribution, as topographic conditions with lower elevations or gentler slopes are favorable for the accumulation of carbon storage, while the decreasing area of carbon storage is more randomly distributed topographically. This study develops a dynamic estimation model of carbon storage considering topographic factors, which provides a solution for the accurate estimation of forest carbon storage in regions with a complex topography.
Southwest China is one of three major forest regions in China and plays an important role in carbon sequestration. Accurate estimations of changes in aboveground biomass are critical for ...understanding forest carbon cycling and promoting climate change mitigation. Southwest China is characterized by complex topographic features and forest canopy structures, complicating methods for mapping aboveground biomass and its dynamics. The integration of continuous Landsat images and national forest inventory data provides an alternative approach to develop a long-term monitoring program of forest aboveground biomass dynamics. This study explores the development of a methodological framework using historical national forest inventory plot data and Landsat TM time-series images. This method was formulated by comparing two parametric methods: Linear Regression for Multiple Independent Variables (MLR), and Partial Least Square Regression (PLSR); and two nonparametric methods: Random Forest (RF) and Gradient Boost Regression Tree (GBRT) based on the state of forest aboveground biomass and change models. The methodological framework mapped
Pinus densata
aboveground biomass and its changes over time in Shangri-la, Yunnan, China. Landsat images and national forest inventory data were acquired for 1987, 1992, 1997, 2002 and 2007. The results show that: (1) correlation and homogeneity texture measures were able to characterize forest canopy structures, aboveground biomass and its dynamics; (2) GBRT and RF predicted
Pinus densata
aboveground biomass and its changes better than PLSR and MLR; (3) GBRT was the most reliable approach in the estimation of aboveground biomass and its changes; and, (4) the aboveground biomass change models showed a promising improvement of prediction accuracy. This study indicates that the combination of GBRT state and change models developed using temporal Landsat and national forest inventory data provides the potential for developing a methodological framework for the long-term mapping and monitoring program of forest aboveground biomass and its changes in Southwest China.