Railway infrastructure inspection is crucial for ensuring operational safety and improving efficiency. However, traditional manual inspection methods are not only time-consuming and labor-intensive but also prone to human errors, leading to fluctuations in inspection quality and efficiency. To address this issue, this paper proposes an unmanned aerial vehicle (UAV) automated inspection method based on a Unified Training Fusion Reinforcement Learning Network (UTFN), which combines Unified Deep Q Networks (DQN) and Proximal Policy Optimization (PPO) algorithms for autonomous route planning and navigation. This approach provides precise support for task path planning in complex geographical environments, overcoming the limitations of traditional methods in such scenarios. The integration of these technologies results in a highly intelligent and automated UAV operation and control inspection system, minimizing human intervention while improving inspection accuracy and efficiency. Experimental results demonstrate that the proposed system effectively reduces operational costs, and enhances operational control efficiency, ensuring the smooth completion of tasks.

In the context of label-efficient learning on video data, the distillation method and the structural design of the teacher-student architecture have a significant impact on knowledge distillation. However, the relationship between these factors has been overlooked in previous research. To address this gap, we propose a new weakly supervised learning framework for knowledge distillation in video classification that is designed to improve the efficiency and accuracy of the student model. Our approach leverages the concept of substage-based learning to distill knowledge based on the combination of student substages and the correlation of corresponding substages. We also employ the progressive cascade training method to address the accuracy loss caused by the large capacity gap between the teacher and the student. Additionally, we propose a pseudo-label optimization strategy to improve the initial data label. To optimize the loss functions of different distillation substages during the training process, we introduce a new loss method based on feature distribution. We conduct extensive experiments on both real and simulated data sets, demonstrating that our proposed approach outperforms existing distillation methods in terms of knowledge distillation for video classification tasks. Our proposed substage-based distillation approach has the potential to inform future research on label-efficient learning for video data.

Graph Convolutional Networks (GCNs) have emerged as a potent tool for learning graph representations, finding applications in a plethora of real-world scenarios. Nevertheless, a significant portion of deep learning research has predominantly concentrated on enhancing model performance via the construction of deeper GCNs. Regrettably, the efficacy of training deep GCNs is marred by two fundamental weaknesses: the inadequacy of conventional methodologies in handling heterogeneous networks, and the exponential surge in model complexity as network depth increases. This, in turn, imposes constraints on their practical utility. To surmount these inherent limitations, we propose an innovative approach named the Wide Sub-stage Graph Convolutional Network (WSSGCN). Our method is an outcome of meticulous observations drawn from classical and graph convolutional networks, aimed at rectifying the constraints associated with traditional GCNs. Our strategy involves the conception of a staged convolutional network framework that mirrors the fundamental tenets of the step-by-step learning process akin to human cognition. This framework prioritizes three distinct forms of consistency: response-based, feature-based, and relationship-based. Our approach involves three tailored convolutional networks capturing node/edge, subgraph, and global features. Additionally, we introduce a novel method to expand graph width for efficient GCN training. Empirical validation on benchmarks highlights WSSGCN’s superior accuracy and faster training versus conventional GCNs. WSSGCN triumphs over traditional GCN constraints, significantly enhancing graph representation learning.

In recent years, the field of distributed deep learning within the Internet of Things (IoT) or the edge has experienced exponential growth. Federated meta-learning has emerged as a significant advancement, enabling collaborative learning among source nodes to establish a global model initialization. This approach allows for optimal performance while necessitating minimal data samples for updating model parameters at the target node. Federated meta-learning has gained increased attention due to its capacity to provide real-time edge intelligence. However, a critical aspect that remains inadequately explored is the recovery of interim meta knowledge’s failure, which constitutes a pivotal key for adapting to new tasks. In this paper, we introduce FMRec, a novel platform designed to offer a fast and flexible recovery mechanism for failed interim meta knowledge in various federated meta-learning scenarios. FMRec serves as a complementary system compatible with different types of federated models and is adaptable to diverse tasks. We present a demonstration of its design and assess its efficiency and reliability through real-world applications.

In the current upgrading and transformation process of smart subway, the ATS system of urban rail transit is faced with such problems as poor stability, cumbersome operation and maintenance methods and low efficiency of fault emergency treatment. Therefore, a micro-service architecture scheme of the ATS system based on the cloud platform is proposed. The overall micro service architecture of the ATS over the cloud platform, the micro-service division and design of ATS applications, container deployment mode and communication interaction mode are studied, which can facilitate loose coupling and easy expansion of the system, save computing resources, improve system stability and operation efficiency.

In order to solve the problems of low quality and low efffciency of CBTC (Communication Based Train Control) system software testing, this article designed a CBTC software intelligent testing system based on big data computing model.This system constructed a multiple linear regression software measurement model, established a reasonable software measurement standard to evaluate the software quality. Based on this, the article proposed an efffcient distributed measurement method based on Hadoop open source framework to solve the big data computing task, andput forward an overall solution of intelligent software testing system to implement the intelligent operation of software testing. Through the experimental analysis, the system can reasonably generate software measurement model, carry out distributed parallel processing of test tasks, reduce the repetitive labor of testers, and improve the quality and efffciency of testers.

An intelligent train testing system based on image recognition technology was designed to solve the problem of automatic simulation of train system test in urban rail transit. The intelligent train test system proposed in this paper adopted the structure model of convolution neural network and convolutional neural network algorithms based on hierarchical compression. The paper introduced in detail the concrete process of constructing layered compression convolution neural network and the optimal structure design of convolution core. Through the analysis of automated simulation experiment and test data of station and yard test cases, the results show that the train intelligent testing system based on convolution neural network optimization algorithm can optimize the test process, reduce manual error operation, rationally allocate test resources, improve test quality, speed up the overall system test schedule requirements. The system can also provide technical support for the implementation of comprehensive automated testing in the ffeld of urban rail transit in the future.

The data mining technology is used to analyze the large data generated during train operation，the decision tree algorithm is proposed based on attribute matrix graph. Combined with the simulation date of a train，the computing attribute matrix and the structure design of the decision tree optimization algorithm are elaborated. According to fault analysis result of the decision tree algorithm，this algorithm could classify the faults accurately and provide reliable basis for the prediction of metro faults.

We introduce the key functions of intelligent car—borne data maintenance terminal software and how to utilize those functions at each stage．The importance and significance of intelligent data analysis in maintenance has also been clarified.

This paper expounds the practical significance of subway train simulation system.By constructing a mathematical model,a complete simulation system of automatic train protection,or ATP,and automatic train operation,or ATO,is built.We focus the research on the construction of simulation model,design of software architecture,modeling calculation and other core issues.

Starting from the interface and function of vehicle DMI in CBTC system,a cross-platform Qt development environment is used for the design and development of vehicle human-machine interface in Linux system and the software flow and interface design are described in detail.Trough lots of tests both in field and in laboratory and constant modifications,the contents of the interface become much richer and the program also can run more stably.