As simultaneous localization and mapping(SLAM)research deepened, the complexity and workload of SLAM tasks increased. Scholars begun to shift their research focus towards multi-robot(or multi-vehicle)SLAM. The collaboration of multiple robots enhanced mapping efficiency. When conducting multi-robot(or multi-vehicle)SLAM, it was necessary to merge local maps to construct a global map. Digital map stitching technology transforms from local maps to global maps through feature matching and fusion of overlapping areas, improving the accuracy and efficiency of map construction. This technology had significant application value in fields such as autonomous driving, multi-robot systems, and geographic information systems. This paper introduced the typical digital maps commonly used in the map stitching process, along with their advantages and disadvantages. It analyzed the factors influencing stitching results, systematically discussing digital map stitching methods around the aspects of homogeneous and heterogeneous map stitching. Additionally, it examined the existing issues in map stitching technology and outlined potential solutions to the challenges posed by heterogeneous map stitching technology.

To tackle the path planning challenges for indoor unmanned vehicle, an improved A^* algorithm and dynamic window approach(DWA)were utilized to develop a hybrid path planning algorithm, which significantly enhanced both global optimality and real-time obstacle avoidance capabilities. Dynamic weights were employed to balance node expansion speed, which boosted the efficiency of the traditional A^* algorithm in complex environments. 24-Neighborhood search strategy was introduced to address the issue of node revisitation in bidirectional searches. The differential in heading angles between successive moments was incorporated into the trajectory evaluation function, optimizing the adaptability of traditional DWA to obstacle distribution, reducing turning angles at obstacles, and increasing travel speed in open areas. An analysis of the planning algorithm's results, supported by simulation experiments, confirmed the efficacy of the hybrid path planning algorithm. Experimental outcomes showed that this enhanced algorithm could effectively ensure optimal path planning alongside robust real-time obstacle avoidance capabilities.

To address the problem of how to reduce the total delay of data transmission with relay between roadside units, considering the adaptive variation characteristics of the V2I downlink data transmission rate within the coverage of roadside units, a data transmission scheduling strategy based on vehicle location information was proposed. This strategy comprehensively considered the randomness of data update and the adaptability of data transmission rate of location difference, and constructed a Markov chain model based on the state transition of a data cache queue. At the same time, considering the existence of multiple vehicles in the roadside unit coverage area, a vehicle priority communication model based on the joint weight of vehicle speed and vehicle position was proposed to determine the communication service strategy. A nonlinear optimization function with the objective of minimizing the total delay of data transmission was established, and the optimal data transmission scheduling strategy of the roadside unit was obtained by linearization. The simulation results showed that the LTS strategy could effectively reduce the total transmission delay under the condition of vehicle arrival rate and data arrival rate change, and the strategy could resist the change of data arrival rate and had good stability of the data transmission.

To further alleviate traffic congestion and improve road network efficiency, this study proposed an urban vehicle route guidance method integrating personalized routing strategies and traffic signal control based on hierarchical multi-agent reinforcement learning(MARL). Route guidance agents and traffic signal control agents were deployed at intersections to provide personalized routing policies and optimize traffic light control, thereby balancing urban traffic flow. To overcome the limitations of predefined graph structures in representing dynamic traffic state features, the traffic signal control agents employed an adaptive graph convolutional network to autonomously capture spatial correlations among peer agents. Concurrently, the route guidance agents integrated meanfield game to analyze aggregated vehicle actions, effectively capturing inter-vehicle interactions for coordinated decision-making while delivering destination-specific routing strategies. To prevent local congestion and severe traffic imbalance, a multi-agent proximal policy optimization(MAPPO)algorithm was adopted, enabling centralized training and decentralized execution for cooperative signal control agents to implement directional flow restriction. A hierarchical reinforcement learning framework facilitated information sharing and collaboration among heterogeneous agents. Extensive experiments were conducted on the SUMO simulation platform using multiple real-world open-source traffic datasets, with comparisons against baseline methods. Results demonstrated that the proposed method reduced average travel time by at least 11.05% and decreased average delay time by at least 19.90%, significantly enhancing urban traffic efficiency.

This study addressed the challenges inherent in the traditional artificial potential field method used in intelligent vehicle path planning, specifically the intricate calibration of the repulsive force parameter, prevalence of local extreme values, imbalance of potential field force, and the issue of unreachable targets. A novel approach was proposed: a dynamic safe elliptical path planning method for intelligent vehicles based on the improved artificial potential field. This study delineated the concept of a transverse virtual safe space, taking into account the interplay between speed and tracking error. It developed an elliptic dynamic safe distance model, both transverse and longitudinal, predicated on dynamic compensation. This model facilitated the calibration of the gravitational repulsive force range, thereby addressing the prevalent challenge of repulsive parameter calibration. Building upon this framework, the paper introduced an equilibrium factor and a variable adjustment factor, both derived from the Gaussian function. These factors were designed considering the relationship between position vectors, effectively addressing issues such as local extremes, potential field force imbalances, and target inaccessibility. A model predictive controller was meticulously designed for the effective tracking control of this path. The experimental outcomes demonstrated the efficacy of the proposed method in resolving the intricate challenges associated with the calibration of repulsive force parameters, local extremes, potential field force imbalances, and target unreachability in intelligent vehicle path planning. Notably, the operational efficiency of the algorithm was enhanced by 69.5% relative to the RRT^* algorithm, and there was a notable 62.2% reduction in the average path curvature. The method significantly enhanced the smoothness and comfort in both single-obstacle and multi-obstacle vehicular simulation planning scenarios, demonstrating its versatility and effectiveness in varying conditions. The real-vehicle experiment results affirmed the applicability of the algorithm in actual path planning scenarios.

By imputing missing data to enhance the quality of traffic flow data, accurate real-time predictions for highway traffic flow can be achieved. This study adopted the low-rank tensor decomposition theory to repair the missing data and used the convolutional long and short term memory neural network based on the attention mechanism to build a prediction model. The actual traffic flow data of the Shandong expressway toll system were used to verify the method proposed in this study. The experimental results showed that: by classifying and identifying missing data, using the confusion matrix to determine the threshold of identifying missing data, introducing the three-dimensional tensor model, and using the low-rank tensor completion method to repair missing data, the data repair effect was the best compared with Lagrange interpolation, KNN, and SVR methods. Compared with SVR, CNN, LSTM, and CNN-LSTM, the combined traffic flow prediction model proposed in this study had higher prediction accuracy in different road sections, working days, and non-working days, and the M_SE and M_APE of traffic prediction decreased by 22.47% and 8.41%, respectively. The average speed forecast for M_SE and M_APE decreased by 42.83% and 6.32%, respectively. The proposed model provided new ideas for research on highway traffic state prediction methods.

This study employed GPR(ground penetrating radar)to identify and locate surface cracks in typical road segments. The method combined core drilling sampling and milling observation for validation, resulting in the construction of a database containing 728 radar images. The YOLO v8l algorithm was used to learn crack features. By incorporating an attention mechanism and modifying the activation function within the YOLO v8l framework, the study overcame the interference caused by the variability of road crack image features and significant noise, while also eliminating model overfitting. After modifying the algorithm, the model's computational parameters increased, and the computational efficiency improved. The precision and recall rates of the revised algorithm reached 99.4% and 92.3%, respectively. During training, the mean average precision and loss function fluctuations were minimal, indicating that the dataset annotation principles were consistent. This proved the effectiveness and reliability of the proposed method for identifying road surface cracks.

To address issues caused by category differences between the pre-training dataset of the object detector and the dataset for the image captioning task, which could lead to object recognition errors, as well as variations in sample sizes across different scenes that could result in the model's insufficient understanding of relationships between objects in rare scenes, the multi-scale visual and textual semantic feature fusion for image captioning(MVTFF-IC)was proposed. The multi-scale visual feature fusion(MVFF)module modeled global, grid, and regional features using a graph attention network to obtain more representative visual representations. The deep semantic fusion module(DSFM)integrated textual semantic features, including object relationships, through a cross-attention mechanism to generate more accurate descriptions. Experimental results on the Microsoft common objects in context(MSCOCO)dataset showed that MVTFF-IC achieved a consensus-based image description evaluation C_DIEr of 136.7, outperforming many popular existing algorithms, demonstrating its ability to capture key information more accurately in images and generate high-quality descriptions.

Aiming at overcoming drawbacks of insufficient search efficiency and tendency to slip into local extremes of beluga optimization algorithm, an enhanced beluga whale optimization(EBWO)algorithm was proposed in this paper. First, a weight-based scramble beluga was included and applied to the algorithm's development phase to enrich the position updating technique, and a greedy mechanism was employed to select a better location and increase the quality of the understanding. Second, an adaptive Gaussian strategy was introduced to locally perturb the beluga in the whale falling phase, to make it adjusted to the vicinity of the optimal position to improve the convergence speed of the algorithm. Finally, a convex lens imaging learning strategy was used to carry out the information position after sharing. The comparative examination of the optimization of the ten benchmark test functions, the CEC2020 test set, and the Wilcoxon rank sum test revealed that EBWO's optimization speed and convergence accuracy had significantly improved. To test the EBWO algorithm's practicality and feasibility, it was applied to solve engineering design problems involving speed reducers and pressure vessels. It was discovered through experimental comparative analysis that the EBWO algorithm had a certain degree of superiority in solving actual optimization problems.

To address the resource allocation problem in multi-user multi-channel heterogeneous cognitive radio-based Internet of Things(CR-IoT)network under the context of spectrum trading, a mathematical analysis method for communication performance was designed, which was developed to meet the need for evaluating the cost-effectiveness of spectrum pricing for multi-service secondary users(SUs)and was applicable to various network environments and operator-specific spectrum allocation strategies. Closed-form expressions for spectrum transaction costs and performance metrics, such as throughput and packet loss rate, were derived. The cost and communication performance of SUs were analyzed independently. A dynamic pricing strategy with customized performance requirements for users was proposed to achieve multi-objective optimization of spectrum trading costs and communication performance. Simulation results demonstrated that the proposed dynamic pricing strategy effectively ensured SUs' quality of service, outperforming fixed pricing strategies and demand-based dynamic pricing strategies.

In order to support geotechnical and rock sample selection in tunnel engineering and rock strength testing in on-site construction, the research used scaled linear cutting test to infer rock strength. A total of 9 batches of rock samples with uniaxial compressive strength of 50-230 MPa were collected from tunnel engineering sites and rock mines. Rod-shaped and block-shaped samples were prepared and subjected to uniaxial compression, Brazilian splitting and scaled linear cutting test respectively; selected 7 batches of test data to obtain the fitting functions of the uniaxial compressive strength, splitting tensile strength and normal force mean, peak value and peak mean value of the scaled linear cutting test, and used the remaining 2 batches of test results to verify the accuracy of the resulting prediction model. The results showed that the fitting correlation coefficients between rock uniaxial compressive strength and normal load were greater than 0.9, and the fitting correlation coefficients between splitting tensile strength and normal load were greater than 0.8, indicating that rock strength had a strong linear correlation with cutting load. The verification test found that when using the average normal force and the average peak value to verify, the error between the predicted value and the experimental value of the uniaxial compressive strength of the rock was less than 5%, and the error between the predicted value and the experimental value of the splitting tensile strength was less than 10%. It showed that the built model had high accuracy. Using the test methods and models proposed in this research, the strength of rock samples could be quickly and accurately estimated, which provided an effective means for rock sampling during indoor testing and rock strength assessment during on-site construction.

In order to study the influence of Yellow River sand on the properties of cement materials,the sand was used to prepare mortar by replacing cement partially. Then, the effect of different sand content on flowability, setting time, and mechanical properties were studied, and the influence of Yellow River sand on cement hydration process, phase composition and microstructure were analyzed through microscopic tests. The results showed that the Yellow River sand shorten the setting time and reduced the fluidity of slurry. The activity index of Yellow River sand on the 28th day was 66.5%, indicating that it played a physical filling role in improving the matrix strength. CaCO₃ in the Yellow River sand provided nucleation sites for cement hydration and accelerated the early hydration process. CaCO₃ also reacted with C₃A to form monocarbonate(MC). And the Ca(OH)₂ participated in the above reaction to form hemicarbonate(HC), which inhibited the transition from ettringite(AFt)to monosulphate(AFm).

In order to investigate the influence of the structural parameters of the reinforced concrete piers reinforced with ultra-high performance concrete(UHPC)on the seismic performance of the piers, based on the results of the proposed static test of the piers, the finite element model of the piers was established by using the ABAQUS. Based on the test data, the accuracy of the finite element model of the pier was verified, and the structural parameters such as the height of the reinforcement layer, the thickness of the reinforcement layer, and the reinforcement of the pier were analyzed using the control variable method. The results of the research showed that with the increase of the height of the reinforced layer UHPC, the plastic damage of the pier body decreased, which positively affected the improvement of the pier body's ability to resist the initial deformation and flexural load capacity. Under a certain thickness of reinforcement layer, the constraining effect of the abutment reached a saturation state, where further increasing the thickness of the UHPC reinforcement layer did not significantly affect the damage to the abutment or any related indicators. The increase in longitudinal reinforcement strength significantly impacted the seismic performance of bridge piers reinforced with UHPC, notably enhancing the pier's horizontal flexural capacity to a greater extent, while concurrently reducing the displacement ductility coefficient of the pier body. The overall development trends of bridge pier skeleton curves were similar for different strengths of longitudinal reinforcement. Substituting equal strength or equal volume with high-strength longitudinal reinforcement delayed the overall stiffness degradation of the pier body, thereby enhancing the elastic working capacity of the bridge pier.

To more accurately predict the deformation of support structures induced by excavation of foundation pits. The research constructed a deformation prediction model for deep foundation pits adjacent to tunnels. The grey wolf optimizer(GWO)algorithm was used to automatically optimize the hyperparameters in the long-short term memory network(LSTM), which improved the accuracy of the original LSTM model's prediction results. Taking a deep excavation project adjacent to a tunnel in the urban area of Nanjing as the engineering background, a comparative analysis was conducted on the prediction results of excavation deformation using BP neural network, original LSTM and GWO-LSTM models. The R2 of the three models were 0.992, 0.967, and 0.999, respectively, indicating the advantages and accuracy of the GWO-LSTM model in predicting deep excavation deformation. Finally, the GWO-LSTM model was used to predict and analyze the deformation of D₁₄ monitoring point and the predicted results were basically consistent with the measured values. The research results could provide technical support for the safe construction of deep foundation pits adjacent to tunnels.

To address curtailment issues caused by large-scale grid integration of distributed photovoltaic systems, a ridge regression-based method for predicting distributed photovoltaic consumption capacity in distribution networks was proposed. Key factors influencing distributed photovoltaic consumption capacity were analyzed, with the contribution degree of these factors quantified through grey relational analysis. A ridge regression-based prediction model was developed by incorporating high-correlation evaluation indicators. The mapping relationships between driving factors and consumption capacity were derived, followed by scenario simulations to formulate strategic recommendations for future photovoltaic absorption improvement. Simulations implemented in the SPSSPRO platform demonstrated that the proposed method accurately predicted photovoltaic consumption capacity, providing actionable insights for enhancing system-level photovoltaic consumption capacity.

Aiming at the problem of insufficient scale of high-quality data samples for data-driven research on theoretical line loss analysis of transformer districts, a calculation method of theoretical line loss in transformer districts based on sample expansion and data-driven was proposed. A generative adversarial network was constructed, and the Adam optimizer was used to optimize the network parameters. The transformer districts samples were analyzed by K-means clustering analysis, while a method for selecting the optimal number of clusters was built based on silhouette coefficient and sum of squared errors. The proper classification of the transformer districts effectively reduced the computational burden of artificial neural network training. The artificial neural network model for theoretical line loss analysis of transformer districts was established through training on each class of transformer districts with the expanded sample set. Simulations were conducted to verify the proposed method by using actual transformer districts data collected from an urban area in Liaocheng City, Shandong Province. The results showed that the sample set was effectively expanded, the training effectiveness of the artificial neural network model was improved, and higher accuracy was achieved in the theoretical line loss analysis of transformer districts.