Abstract:

Abstract: This study designed a goal-directed training system to conduct video learning and ob＜x＞ject transfer tests on pigeons. The behavioral experiments consisted of three phases: pigeon screening and adaptation, video learning, and ob＜x＞ject transfer testing. Electrodes were then implanted in the MVL nucleus to record neural signals during the recognition of video and ob＜x＞ject targets. The power spectrum before and after target recognition was calculated using the Welch method, identifying the characteristic frequency bands associated with target recognition in the pigeons MVL brain region. Finally, a brain functional network was constructed ba＜x＞sed on phase-locking value (PLV), and features such as average node degree, clustering coefficient, and global efficiency were extracted. The behavioral results showed that video learning significantly improved the pigeons ability to recognize ob＜x＞jects, confirming the brains capability for transfer learning in visual cognition. The analysis of brain functional network features in the MVL nucleus during visual target recognition suggested that it plays a crucial role in both visual target recognition and visual transfer learning. The MVL, as a higher-order visual nucleus, may extract common features between video and physical ob＜x＞jects in visual target recognition.

Abstract: To address the problem of getting stuck in local optimal solutions in UAV path planning with particle swarm algorithm and insufficient consideration for obstacle avoidance after smoothing discrete path points, a threedimensional B-spline curve path planning method for unmanned aerial vehicles based on an improved particle swarm algorithm was proposed. Firstly, considering the flight performance requirements such as UAV path length, safe obstacle avoidance, flight altitude, and smoothness, a path planning model was constructed using the geometric properties of B-spline curves. Then, an improved particle swarm algorithm was used to solve the model. The algorithm improvement was mainly achieved by optimizing the particle initialization strategy, updating the inertia weight factor and learning factor strategy, and increasing the particle perturbation strategy. The test results on the CEC2017 standard test function set showed that the improved particle swarm algorithm exhibited stronger optimization ability and better stability compared to other algorithms. The simulation results of two scenarios showed that the planned path cost was reduced by 2%, stability was improved by 65%, path safety avoided obstacles and C 2 was continuous, which could meet the comprehensive performance requirements of UAV flight.

Abstract: Aiming at the problems such as lack of infrastructure, high task delay and high bandwidth demand in complex geographical conditions, a multi-stage mobile edge computing system model which combined computing offloading and power distribution was proposed. In this model, a server equipped with MEC was deployed near the UAV to provide computing services, and the problems such as task offloading, power consumption and computing resource allocation of the UAV were comprehensively analyzed and the measurement methods were given. At the same time, the types of tasks that the UAV could perform and the requirements of the CPU and GPU on the UAV were considered. The problem was expressed as a mixed integer nonlinear problem. A task computing offloading algorithm based on deep reinforcement learning was proposed to solve this problem. Based on the improved double deep Q learning algorithm, the algorithm used deep neural network to find the mapping between UAVs in deep reinforcement learning, finding potential patterns from the state space and estimating the optimal action, and used model-free DRL method to enable each UAV to make quick offloading decisions based on local observations. Simulation results showed that the proposed algorithm reduced the average offloading cost by 42. 8% compared with LCGP algorithm. Compared with DDPG algorithm, the energy consumption was reduced by 16%. Compared with DDQN algorithm, the task execution delay was reduced by 12. 9%.

Abstract: In order to solve the problem that the tracking accuracy of agricultural machinery vehicles on the reference path was not easily influenced by unknown disturbances and could not meet needs of various operating environments, the most common front-wheel steering and rear-wheel driven wheeled agricultural machinery vehicles infarmland operation was taken as the object, and the tracking control problem of agricultural machinery vehicles withthe fully actuated control approaches was studied. The widely used proportional-integral-derivative ( PID) controland active disturbance rejection control were used as comparison controllers. And the controllers designed with theproposed control method were compared and simulated in different experimental environments. The results showedthat, compared with the other two methods, the proposed method performed better in tracking the reference path,more stable in different experimental conditions, and with higher control accuracy and robustness.

Abstract: To address the problem of target identity ( ID) fluctuation during target tracking, which might affect the time-sensitive target recognition, an " detection-decision" time-sensitive target recognition method ( AR-SSVEP-YOLOv3) was proposed which integrated augmented reality (AR) technology, steady state visual evoked potential ( SSVEP) , and YOLOv3. The target perception module obtained the front-end scene video and presented it in reatime through an AR headset. The YOLOv3 algorithm completed the detection of sensitive targets in the scene video, and the AR-SSVEP EEG processing module decoded the EEG data of the subject during ID changes to identify time-sensitive targets. The correct recognition rate of time-sensitive targets was compared and analyzed. The results showed that the average improvement was 18. 8% in the recognition accuracy of AR-SSVEP-YOLOv3 time-sensitive target recognition method compared with the YOLOv3 algorithm, and the average improvement was 8. 0% compared with the YOLOv3-Sort algorithm. The AR-SSVEP-YOLOv3 time-sensitive target recognition method could reduce the influence of target ID fluctuation on time-sensitive target recognition and improve the human-computer interaction ability and the correct recognition rate of time-sensitive targets.

Abstract: An abandoned object detection algorithm based on improved YOLOv8 was proposed to address the difficulties of traditional background subtraction based abandoned object detection algorithms in dealing with crowded environments, small targets, occlusion, and light changes, as well as the low accuracy of models based on deep learning methods. Firstly, dynamic upsampling DySample was used to replace the nearest neighbor upsampling, optimizing the upsampling process, and increasing the model′s generalization ability. Secondly, the downsampling convolution was replaced with the efficient lightweight ADown module which reduced the overall model parameters while improving the detection accuracy of the algorithm. In addition, the introduction of EMA attention mechanism optimized the feature extraction process, enhanced feature extraction capabilities, and improved the effectiveness of small object detection. The experimental results showed that the improved model YOLO-DAE achieved P, R, and mAP@ 50 and mAP@ 50:95 was 93. 4%, 87. 7%, 91. 7%, and 80. 2%, respectively, which was 1. 8, 1. 6, 1. 2, and 2. 1 percentage points higher than the original YOLOv8s. And the average accuracy mAP@ 50 and mAP@ 50: 95 was higher than YOLOv5s r6. 0, YOLOv6s v3. 0, YOLOv7s AF, and YOLOv9s, effectively improving the ability to detect abandoned object。

Abstract: Distributed hybrid flowline rescheduling was investigated considering machine breakdown and transportation time constraints. An integer programming model was constructed with the optimization objective of simultaneously minimizing the maximum completion time, total energy consumption, and total delay. An improved grey wolf optimization algorithm was then proposed to solve it. Firstly, according to the characteristics of the problem, a three-chain encoding method based on factory-operation-machine was designed. A population initialization method combined with NEH heuristic approach and completely random procedure was proposed. Next, after the leadership individuals were chosen, a dual-mode parallel search method based on tracking and autonomous action was introduced to update the bottom wolves. Finally, tabu search integrated with forward insertion transformation of operation chain and backward shift operation of machine chain was applied to avoid falling into local optimum. Simulation experiments tested 370 instances. The effectiveness of the improvement items in the proposed algorithm was verified. The improved grey wolf optimization algorithm improved by 9. 33%, 12. 24%, 10. 43%, and 9. 61%, respectively, compared with the four algorithms, including hybrid beluga whale optimization algorithm, hybrid flower pollination algorithm, classical grey wolf optimization algorithm and improved moth-flame optimization algorithm. It illustrated the effectiveness of the proposed algorithm.

Abstract: Aiming to meet the requirement of modulation recognition in high order and to solve the difficulty of modulation recognition in low signal-to-noise ratio environment in 6G communication, a modulation recognition algorithm based on multi-criteria fusion and intelligent decision was proposed by combining artificial intelligence technology and modern signal processing technology. The algorithm was divided into two parts. Multi-criteria fusion network and intelligent decision network. The multi-criteria fusion network calculated the higher-order cumulative extensions of the standard modulation signals, traversed all the potential thresholds by using local optimal solutions, and determined the judgment thresholds by Gini coefficient and the entropy of certainty gain. The intelligent decision network adopted a CART architecture to recognize the modulation format of unknown signals using the determined judgment thresholds, and the model was iterative optimized using a pruning algorithm to obtain the finally optimal decision tree, forming a modulation recognition algorithm based on multi-criteria fusion and intelligent decision making. Experimental results showed that the algorithm could accurately recognize 16QAM, 64QAM, 128QAM, 1024QAM, 2PSK, 4PSK, 8PSK, 2FSK, 4FSK at 0 dB SNR, and the comprehensive recognition accuracy reached 99. 4%. Compared with other methods, the modulation recognition accuracy and the types of recognizable modulation were improved.

Abstract: Aiming at the problems of lack of small target information during feature extraction, partial loss of information during feature fusion, and inconspicuous small target feature information in remote sensing image target detection task, which lead to the low accuracy of small target detection, an algorithm for remote sensing image target detection based on FEW-YOLOv8 model was proposed. Firstly, the backbone network architecture was optimized to use the FasterNet backbone network, which extracted the spatial features of small targets in remote sensing images more efficiently, making the network model more focused on tiny targets, thus improving the small target detection accuracy. Secondly, the new C2f_EMA module was constructed using EMA attention and C2f to replace the C2f module in Neck network, and the feature attention enhancement operation was performed before fusing the features, so that the network model highlighted the small-target part of the feature information more, which effectively solved the problem of small-target feature loss in the process of feature fusion. Finally, WIoUv3, which had a dynamic non-monotonic FM, was used as the bounding box loss function to improve the accuracy of the model′s bounding box localization and strengthen the localization ability of small targets. The experimental results on NWPU VHR-10, HRSC2016 and DOTA v1. 0 datasets showed that the test mAP50 of the improved YOLOv8 algorithm was 7. 71, 9. 70 and 12. 32 percentage points higher than that of the original YOLOv8 algorithm, respectively, which proved that the proposed algorithm could effectively improve the detection accuracy of small targets in remote sensing images.

Abstract: To explore the inherent characteristics of Tai Chi Stake training and digitally interpret its movement essentials, an effective plantar pressure detection equipment was used to collect plantar pressure signals from participants, and the relative position, time-domain, frequency-domain, and regularity indicators of the center of pressure (COP) movement between the expert group and the trainee group were compared and analyzed. The results showed that the relative position of COP of the expert group was closer to 50% compared to the trainee group. In terms of time-domain indicators, the root mean square of COP movement of the expert group was significantly larger than that of the trainee group in both the left-right and front-back directions, while the frequency-domain peak frequency of the movement was significantly lower than that of the trainee group in both directions. For the sample entropy analysis of COP that measured the regularity, the expert group showed significantly lower values in both the left-right and front-back directions compared to the trainee group. The Tai Chi Stake COP of the expert group was more concentrated in the central position, reflecting the technical essential of " stand straight and be centered" . The regular low-frequency adjustment reflected the characteristic of " motion in quiescence" in Tai Chi Stake.

Abstract: Given the challenge of limited high-resolution human contact data during the early stages of emerging infectious disease outbreaks, it is difficult to implement early warning strategies via the global structural characteristics of contact networks. Multi-source data-driven sentinel surveillance strategies for infectious diseases were the focuses of this study, and a novel framework for emerging infectious disease risk surveillance based on urban contact networks was proposed. By integrating multi-source census and survey data, a contact network reflecting the characteristics of the urban population structure was constructed to simulate the transmission of emerging infectious disease in specific cities. Based on this, a " one person per household" surveillance strategy was proposed. This strategy leveraged a small number of selected sentinel samples to achieve near-whole population coverage for effective risk surveillance, eliminating the need for prior knowledge of the global network structure. Experimental results demonstrated that during periods of low disease transmissibility ( basic reproduction number of 1. 2) , the proposed household surveillance strategy performed at the same level to the random surveillance strategy, while with lower cost compared with surveillance the whole population. As transmissibility increased ( basic reproduction number from 2. 0 to 3. 0) , the early warning performance of household surveillance strategy ranked the second only to the most connected strategy, effectively capturing the transmission of emerging infectious diseases. Notably, it effectively captured the transmission risk of emerging infectious diseases, providing an early warning time of 1. 03 d(37%) and 0. 69 d(53%) compared with the random surveillance strategy.

Abstract: In response to the issue of traditional incentive mechanisms that required users to disclose personal value judgments in advance, potentially leading to privacy leakage, a mathematical model for mobile crowdsensing was established, to clarify key factors such as sensing tasks, value functions, budgets, and user benefits. Then, an MCCA mechanism based on clock auction was proposed to effectively address privacy leakage. The mechanism consisted of an initial allocation pricing phase and a final winner determination phase. Both could effectively protect uner privacy. Theoretical analysis demonstrated that the MCCA algorithm satisfied all requirements of truthfulness, individual rationality, budget feasibility, and efficiency. In the experimental section, MCCA was compared with existing algorithms from the perspectives of use scale, budget scale, and POI scale. The results showed that MCCA achieved comparable value gain to existing algorithms while significantly improving execution efficiency and successfully preventing user privacy leakage.

Abstract: In order to solve the problem of low efficiency of steel defect detection and economic loss caused by false detection, an improved YOLOv5 algorithm was proposed for steel defect detection. On the condition of keeping the original YOLOv5 detection layer unchanged, three auxiliary branches with adaptive weights to extract the shallow information of the YOLOv5 network were added to the improved algortihm, and the auxiliary branches could also enhance the gradient flow of the whole network, which made the training effect better. The EMA attention mechanism was added to the main part of the network, and the weighted feature information of the EMA module could help the model better focus on and understand the important target features. SIoU was used instead of the CIoU loss function, and the angle loss and shape loss introduced by SIoU could make the anchor frame fast and accurate in the regression process to improve the stability and robustness of the detection. Through experiments on the NEUDET dataset, the proposed algorithm improved the accuracy by 3. 7 percentage points compared with the original YOLOv5s, with better detection performance than other mainstream algorithms.

Abstract: The deep soil-bearing heavy ice layer is the extreme adverse geology, on which the construction of roadbed will be exposed to serious frost heave, and thawing and settlement disease, threatening the safe operation of vehicles. Polyurethane polymer ( PU) material was investigated as a thermal insulation layer for roadbeds to prevent freeze-thaw damage. The thermal insulation capability test of PU was conducted to analyze the effect of density and number of freeze-thaw cycles on the thermal conductivity of PU. The model test of thermal insulation of heavy ice layer frozen soil roadbed was carried out. The temperature distribution characteristics of ordinary roadbed, single-layer PU board roadbed and double-layer PU board roadbed during freeze-thaw process were investigated and the thermal insulation effect of PU board was described quantitatively. The results showed that the thermal conductivity of PU was positively correlated with its own density and the number of freeze-thaw cycles. The lower the density of PU, the more its thermal conductivity was affected by freeze-thaw cycles. The higher the density of PU, the more its thermal insulation performance could remain stable in multiple freeze-thaw cycles. The heat flux of ordinary roadbed was 1. 7 times of single-layer PU board roadbed and double-layer PU board roadbed in the freezing process, and 2. 1 times of single-layer PU board roadbed and 2. 8 times of double-layer PU board roadbed in the thawing process. PU board had thermal insulation ability, which could lift the freezing depth of the roadbed and reduce frost heave disease. The existence of PU could also prolong the freezing process of the roadbed and avoid thawing and settlement disease. Double-layer PU board showed better thermal insulation effect than single-layer PU board.

Abstract: In view of the complex design process of the spiral case with cushion layer, the cumbersome nature of design changes, and the low efficiency of structure optimization, design parameters for the spiral case with cushion layer were established based on the Inventor platform, and the parametric modeling method of the spiral case was investigated. To tackle the challenge of handling contact states between the steel lining and cushion layer, as well as between the cushion layer and outer concrete during parametric modeling, a surface partitioning algorithm was introduced, enabling effective segmentation among the steel lining, cushion layer, and outer concrete. For the overlap between spiral case with cushion layer and the inlet cushion layer, an entity contact determination algorithm and cutting method were proposed, achieving refined treatment at their intersections. Case studies showed that using the proposed parametric design approach allowed for rapid model creation based on design parameter adjustments. As the thickness increased, the proportion of internal water pressure borne by the outer concrete decreased, reaching a minimum of 20. 16%, while the circumferential and radial displacements of the steel lining increased, with maximum values of 12. 26 mm and 9. 39 mm, respectively. As the laying range expanded, the proportion of internal water pressure borne by the outer concrete roughly showed a decreasing trend, with a minimum of 23. 17%, while the circumferential and radial displacements of the steel lining decreased, with minimum values of 4. 27 mm and 2. 06 mm, respectively. Ultimately, the meridional wrapping angle of the cushion layer was selected to extend 15° below the waistline, with a thickness of 20 mm.

Abstract: Aiming at the problem that the potential life benefits of river treatment projects are ignored due to the difficulty in quantifying the value of life, a quantification method for the potential life benefits in the flood control benefits of river treatment projects was proposed. Combined with the loss of life caused by floods, the income elasticity coefficient method was introduced to calculate the statistical value of life, and a quantification function for the monetary value of the loss of life caused by floods was constructed. The frequency method for calculating the flood control benefits of water conservancy projects was simplified, and a quantification method for the potential life benefits of river treatment projects based on the statistical value of life was proposed. The method was applied to the comprehensive treatment project of the Jialu River. The results showed that the average potential life benefits of the project over many years was 30. 735 4 million yuan / a. Among them, Zhongmu County accounted for 39. 5%, Chuanhui District of Zhoukou City, Xihua County, Fugou County and the urban area of Weishi County together accounted for 38. 7%, and the non-urban areas along the line accounted for 21. 8%. The potential life benefits of river treatment projects are mainly in urban areas and densely populated areas. By quantifying the potential life benefits in flood control benefits, the calculation theory and method of flood control benefits of water conservancy projects were improved.

Abstract: The existing compaction grouting simulation method only can analyze the stress distribution after grouting, but cannot obtain the parameter information reflecting the compaction effect of grouting, such as the void ratio and density of soil after compaction. Therefore, a modified Cam-clay ( MCC ) model was introduced to describe the mechanical property of soil, and based on the elastic-plastic finite element theory, a simulation method was established to simulate the compaction grouting process of constant density slurry in soil. A more comprehensive and intuitive description of the formation compaction effect was achieved. The compaction grouting simulation analysis was carried out on clay, silty clay and other low permeability soil. Compared with the analytical solutioin and experimental results, the overall average relative errors of the simulated values and analytical solutions of radial stress and void ratio with different grouting pressures were 4. 04% and 0. 29%, respectively, and the average relative errors between the calculated void ratio and elastic modulus and the field test results were 5. 70% and 2. 85%, respectively, which proved the applicability of the proposed model. On this basis, the distribution characteristics of soil density, void ratio and elastic modulus around the grouting column after grouting reinforcement were analyzed. The results showed that when the grouting pressure increased from 0. 4 MPa to 1. 0 MPa at the grouting depth of 1. 5 m, the soil density, elastic modulus and void ratio at 0. 05 m from the center of the grouting hole approximated linear changes, and the average change rates werre 0. 148 g / ( cm 3·MPa) , 0. 808 and -0. 127 MPa - 1 , respectively. When the grouting pressure was 0. 4 MPa, the increase rates of soil density and elastic modulus and the decrease rate of void ratio at 0. 05 m from the center of the grouting hole gradually decreased with the increase of grouting depth. Overall, the density and elastic modulus of the soil around the grouting column were greatly increased after grouting reinforcement, and the void ratio was significantly reduced. The soil parameters changed less with the distance from the grouting hole, and gradually returned to the initial state. On the same grouting pressure condition, with the increase of grouting depth, the compaction effect gradually weakened.

Abstract: A short-term wind power prediction model based on POTDBO-VMD-CNN-BiLSTM was proposed to improve the accuracy of short-term wind power prediction. Firstly, three strategies were adopted to improve the dung beetle optimization algorithm, including integrating Piecewise chaotic mapping, integrating Osprey optimization algorithm, and integrating adaptive T-distribution perturbation, in order to balance the global exploration and local development capabilities of the dung beetle optimization algorithm and accelerate its convergence speed. Secondly, the improved dung beetle optimization algorithm ( POTDBO) was used to optimize the decomposition number and penalty factor of variational mode decomposition (VMD) to improve the decomposition effect of VMD. Then, the POTDBO-VMD model was used to decompose the wind power. Finally, the decomposed frequency components and residual components were input into the CNN-BiLSTM hybrid model for prediction, and the prediction results of each frequency component and residual component were sequentially reconstructed to obtain the wind power prediction results. The proposed model was experimentally tested using actual data from wind farms in Xinjiang and Jilin. Compared with the CNN-BiLSTM model, the results showed that the proposed model increased by 4. 21% and 7. 69% on R 2 respectively, demonstrating better prediction accuracy.

Abstract: In order to make effective use of the pressure energy in the transmission and regulation process of natural gas networks, a integrated energy system scheme involving the comprehensive use of power generation and cold energy of the natural gas pressure was proposed. Firstly, considering that natural gas pressure energy could be used with power generation and refrigeration, a model of electricity-heat-gas-cold integrated energy system containing natural gas pressure energy was established. Secondly, an economic optimization scheduling model with the minimum daily operating cost as the objective function was proposed including the cost of power purchase, gas purchase and equipment operation and maintenance, etc. Finally, the mixed-integer nonlinear optimization model was solved based on MATLAB platform combined with CPLEX solver. The economy and effectiveness of the proposed model were verified with the operation data of a real industrial park. The results showed that the operating cost of the system could be reduced by 74. 9% compared with no natural gas pressure energy utilization, and the system could obtain a good economic benefit.