PaperlayerView exact revision on arXiv

Source-linked AI summary

6G-Enabled Smart Railways

Bo Ai, Yunlong Lu, Yuguang Fang, Dusit Niyato, Ruisi He, Wei Chen, Jiayi Zhang, Guoyu Ma, Yong Niu, Zhangdui Zhong

arXiv:2505.12946v1eess.SYcs.SY

TL;DR

Future smart railways require capabilities that 5G cannot fully meet, including high speed, low latency, security, coverage, and positioning accuracy. This review examines an integrated 6G railway architecture and enabling technologies. It concludes that these technologies have potential but require railway-tailored solutions and further development.

  • Problem

    Future railway requirements for speed, latency, security, coverage, and positioning exceed what 5G can fully meet, while massive IoT access faces limited spectrum.

    Source: A. OTFS in Smart High-Speed Railway · p. 17, §VIII · p. 37

  • Method

    The review examines integrated railway network architecture and enabling technologies, including channel modeling, coverage, communications-computing integration, and digital twins.

    Source: F. Computing and Caching · p. 6, §VIII · p. 37

  • Results

    The review identifies integrated architecture and railway-tailored solutions as central to applying 6G technologies in smart railways.

    Source: §VIII · p. 37

  • Takeaways & Limitations

    Future development should address railway-specific challenges in deploying and innovating the reviewed technologies.

    Source: §VIII · p. 37

Abstract

from arXiv · show

Smart railways integrate advanced information technologies into railway operating systems to improve efficiency and reliability. Although the development of 5G has enhanced railway services, future smart railways require ultra-high speeds, ultra-low latency, ultra-high security, full coverage, and ultra-high positioning accuracy, which 5G cannot fully meet. Therefore, 6G is envisioned to provide green and efficient all-day operations, strong information security, fully automatic driving, and low-cost intelligent maintenance. To achieve these requirements, we propose an integrated network architecture leveraging communications, computing, edge intelligence, and caching in railway systems. We have conducted in-depth investigations on key enabling technologies for reliable transmissions and wireless coverage. For high-speed mobile scenarios, we propose an AI-enabled cross-domain channel modeling and orthogonal time-frequency space-time spread multiple access mechanism to alleviate the conflict between limited spectrum availability and massive user access. The roles of blockchain, edge intelligence, and privacy technologies in endogenously secure rail communications are also evaluated. We further explore the application of emerging paradigms such as integrated sensing and communications, AI-assisted Internet of Things, semantic communications, and digital twin networks for railway maintenance, monitoring, prediction, and accident warning. Finally, possible future research and development directions are discussed.

I. INTRODUCTION

Railway communication underpins smart-railway integration, but 5G-R is projected to fall short of future high-speed, high-capacity and safety-related service demands. The paper reviews 6G requirements, architectures, enabling technologies, and research directions.

A. Development Progress of 5G Smart Rail Communications

  • 5G railway services already support broadband access, monitoring, remote operation, and applications including autonomous driving and railway IoT.The 3GPP, ITU, ETSI, UIC, and industry groups are also working on future railway communication specifications and deployment.Source: A. Development Progress of 5G Smart… · p. 1
  • 5G indicators support 500 km/h, while some terminals may exceed 1000 km/h; 5G also lacks defined safety and positioning-accuracy indicators.Future services are expected to need 10–50 times improvements in latency and reliability over 5G.Source: A. Development Progress of 5G Smart… · pp. 1–2

B. Design Challenges and Visions for 6G Smart Railways

  • 6G railway visions include all-day automatic operations, lifecycle-wide intelligent maintenance, centimeter-level train positioning, and real-time digital-twin monitoring and alerting.Challenges include high-speed mobility, continuous coverage across remote areas, and real-time data exchange among trains, control centers, and signaling systems.Source: B. Design Challenges and Visions for… · p. 2

C. Comparisons and Key Contributions

  • Earlier railway surveys address topics such as heterogeneous networks, railway-system technologies, 5G NR, QoS, or IoT, but lack a comprehensive integrated 6G treatment.Some also do not address active perception, optimal resource allocation, or smart-railway-specific characteristics.Source: C. Comparisons and Key Contributions · p. 2
  • The paper reviews 5G railway communications, 6G requirements, enabling technologies, and deployment challenges, including channel modeling, coverage, edge intelligence, and endogenous security.It also presents deployment methods for technologies such as cell-free MIMO, RIS, and digital twins.Source: C. Comparisons and Key Contributions · pp. 2–3
  • The review also covers integrated communications, computing, and caching with active sensing and resource allocation, and discusses future research directions.Its scope includes IoT monitoring and transmission, massive machine communications, and integrated sensing and communication.Source: C. Comparisons and Key Contributions · pp. 2–3

A. Network Architecture

The section surveys an integrated 6G railway architecture and enabling technologies, emphasizing that railway mobility, environments, and coverage needs require tailored designs.

A. Network Architecture

  • The proposed railway architecture integrates security, AI, and digital twins into a space-air-ground network, while holistic service-based architecture extends functions across control, user, and access domains.The logical architecture is organized into three bodies, four layers, and five planes.Source: A. Network Architecture · p. 3, Figure 1 · p. 3

B. Wireless Channels

  • Railway channel modeling must span diverse domains, including space-ground, RIS, mmWave, UAV, and IoT channels, with rapidly varying V2V and high-mobility propagation conditions.V2V dual-train movement complicates Doppler and channel stationarity; mmWave faces propagation loss and stronger Doppler effects.Source: B. Wireless Channels · p. 4
  • Air-to-ground railway channel models remain limited by a lack of measurements in representative rail scenarios, while RIS channel research still has gaps in propagation, modeling, and high-speed wave control.The text calls for measurement campaigns and accurate high-mobility air-to-ground models, and identifies RIS modeling and wave-control questions as unresolved.Source: B. Wireless Channels · p. 5

C. Wireless Coverage

  • Reliable railway coverage is constrained by high mobility, complex environments, large time-space-frequency networks, heterogeneous services, and limited resources.Handover failures and interruptions are concerns in fast time-varying channels and diverse railway terrain.Source: C. Wireless Coverage · p. 5

D. Cell-free Massive MIMO

  • Cell-free massive MIMO uses distributed access points connected to a central processing unit and shared time-frequency resources to provide coverage without conventional cell-edge effects.For high-speed rail, it may mitigate Doppler-related inter-subcarrier interference, but scalable deployment and severe Doppler remain challenges.Source: D. Cell-free Massive MIMO · p. 6

E. RIS in High-Speed Railway

  • A RIS uses passive reflecting elements to adjust incident-wave amplitude or phase; in high-speed rail, virtual line-of-sight paths can mitigate occlusion and expand coverage.RIS reconfiguration can also counter Doppler and fading and may reduce handovers, hardware costs, and energy consumption.Source: E. RIS in High-Speed Railway · p. 6

F. Computing and Caching

The section relates railway computing, security, and emerging technologies to the demands of 6G operations. It highlights railway-specific challenges and the need to tailor technologies to them.

F. Computing and Caching

  • Railway computing and caching must account for predictable track-bound mobility, safety-critical low-latency services, periodic handovers, tunnel shadowing, and heterogeneous networks.The section also identifies proactive location-aware caching, fault-tolerant mechanisms, and concurrent offloading for massive terminals as needs.Source: F. Computing and Caching · p. 6

G. Security and Privacy

  • Railway security concerns span cloud platforms, train control, edge networks, and massive-access terminals, motivating endogenous security and blockchain-based trust management.Discussed technologies include spatiotemporal trust matching, authentication, fast switching, intrusion detection, and data privacy protection.Source: G. Security and Privacy · pp. 6–7

1) Ubiquitous AI:

4) Semantic Communication:

  • Semantic communication transmits essential contextual information rather than full data streams, a potential way to reduce bandwidth use for railway video surveillance in constrained networks.The passage also identifies autonomous driving, intelligent scheduling, and passenger services as application areas.Source: 4) Semantic Communication · p. 7

5) Cooperative Secure Transmission:

  • The security overview identifies physical-layer, quantum-safe, AI-driven, and trust-establishment technologies, including blockchain and distributed ledgers for confidentiality and integrity.Open authentication protocols are also discussed for integrated networks.Source: 5) Cooperative Secure Transmission · pp. 7–8

6) Digital Twin:

  • Digital twins support virtual-physical modeling, dynamic control, predictive operations, and iterative optimization; for railways, they are discussed for monitoring and reducing operation and maintenance costs.Prior work cited in the section includes DT-assisted offloading and adaptation to network dynamics.Source: 6) Digital Twin · p. 8

I. Lessons Learned

  • The lessons learned identify integrated architecture, tailored channel and coverage solutions, AI, digital twins, and semantic communications as important, while calling for railway-specific technology development.The broader architecture section presents a flexible framework for integrating enabling technologies.Source: §I · p. 8, §III · p. 8

A. Integrated 6G Network Architecture

The proposed railway network combines communications, computing, storage, and intelligence to address connectivity, intelligent-access, and security challenges. Its architecture spans user, access, and core layers and terrestrial, aerial, and space networks, while channel modeling must account for railway-specific mobility and propagation conditions.

A. Integrated 6G Network Architecture

  • The network is intended to combine diverse communications with distributed intelligence and endogenous security across varied terrain and operating conditions.The design targets full connectivity, massive intelligent connections, and secure, reliable automated railway operations.Source: A. Integrated 6G Network Architecture · p. 8
  • The architecture has user, access, and core layers, with edge resources supporting on-demand connections and real-time computation and core resources handling tasks beyond edge capacity.The user layer connects trains, mobile devices, trackside equipment, and IoT sensors; the core provides large-scale computing.Source: A. Integrated 6G Network Architecture · pp. 8–9
  • The network divides into terrestrial, aerial, and space segments, using trackside links, airships or UAVs, and satellites for communication and coverage.Distributed AI in the aerial segment can support cooperative learning, event prediction, and operation scheduling; satellite links serve remote areas and tunnels.Source: A. Integrated 6G Network Architecture · p. 9

B. Cross-Domain Channel Modeling

  • Accurate channel characterization and modeling underpin the design, evaluation, and standardization of dedicated railway communications used for safety-related services.The cited services include dispatching commands, train operation control, and early fault warning.Source: B. Cross-Domain Channel Modeling · p. 9

1) Train-to-Train Channel:

  • Train-to-train channels face fast fading, Doppler-driven non-stationarity, frequent NLOS links, and complex multipath, motivating targeted high-speed measurements and multidimensional modeling.Existing measurement coverage is insufficient, especially for high-speed, multi-frequency T2T links; simulation models still need measured-data calibration and verification.Source: 1) Train-to-Train Channel · p. 10

2) Space/Air-to-Ground Channels for Railways:

  • Space/air-ground networks can extend high-speed railway coverage to remote areas, but channel modeling and deployment must account for obstacles, weather, mobility, delay, and propagation loss.UAV path planning must balance coverage and collision avoidance; satellite links also face propagation and handover characteristics that vary by orbit.Source: 2) Space/Air-to-Ground Channels for… · pp. 10–11

3) Railway Channels at High Frequency Bands:

Railway channels must support both reliable train-control links and high-capacity passenger or hotspot services across high-mobility, complex environments. The section discusses frequency-specific channel needs and an onboard RIS model whose optimized phase produces stronger, less variable CIR as RIS element count increases.

3) Railway Channels at High Frequency Bands:

  • High-frequency railway channels remain an important research direction because mmWave systems must operate amid high mobility, frequent handovers, and changing surroundings.The passage identifies mmWave measurement and modeling in HSR scenarios as an urgent study direction.Source: 3) Railway Channels at High Frequency… · p. 11

4) Train-to-Ground Communications:

  • Train-to-ground systems must carry reliable train-control information and high-rate passenger entertainment, with low-frequency bands suited to control and mmWave to high-rate services.Most existing HSR investigations focus on sub-6 GHz; more mmWave measurement campaigns are still required.Source: 4) Train-to-Ground Communications · pp. 11–12

5) Railway Hotspot Communications:

  • Railway hotspots have many users and heavy traffic but slow mobility, making mmWave and sub-THz candidate technologies for large-capacity services.THz access points are described as offering at least 20 Gbps, while deployment remains challenged by ultra-high-frequency device performance.Source: 5) Railway Hotspot Communications · p. 12

6) RIS Channels for Railway:

  • Existing RIS channel studies largely cover static settings or general RIS models, leaving high-speed train scenarios underrepresented.The section notes that RIS-assisted high-speed train channel modeling and characterization are crucial, while real-world validation is difficult because HSR experiments have limited access and mmWave RIS size is constrained.Source: 6) RIS Channels for Railway · pp. 12–13
  • Onboard RIS motivates a 3D cluster-based model that separates BS-RIS-Rx and BS-Rx links and represents five propagation cases.The cases are single or multiple bounces at the RIS, LoS, single bounce, and multiple bounce; onboard placement also makes the RIS-to-receiver channel quasi-static.Source: 6) RIS Channels for Railway · p. 13
  • More RIS elements increase CIR magnitude and decrease its variations under the RIS optimization phase.The passage attributes this to a larger proportion of SBR components, which weakens the influence of other multipath components.Source: 6) RIS Channels for Railway · p. 13, Figure 5 · p. 13

7) Railway IoT Channels:

Railway IoT channel models remain underexplored across ground and intra-train links, while AI-based modeling offers a satellite-image approach to predicting path loss. The proposed model improves test-set RMSE over a receiver-area-image baseline and shows lower maximum street-level error.

7) Railway IoT Channels:

  • Railway IoT includes ground cellular links and intra-train D2D links, but its propagation models need further study across path loss, shadow fading, small-scale fading, and mmWave/THz channels.Existing work includes NB-IoT path-loss modeling and D2D models with path-loss exponents of 4.13–4.80 and shadow-fading standard deviations of 8.87–10.96 dB.Source: 7) Railway IoT Channels · pp. 13–14

8) AI Enabled Railway Channel Modeling:

  • Conventional statistical and deterministic channel models respectively lack geographic detail or require costly environmental data and computation, motivating generalizable, efficient, accurate models for changing HSR environments.Environmental reconstruction using sensors or ray tracing is costly and computationally complex for real-time railway requirements.Source: 8) AI Enabled Railway Channel Modeling · p. 14
  • The proposed model combines satellite-image features from a CNN with system-state features from an MLP to predict path loss.Inputs include transmit power, antenna height, and transceiver distance alongside the image; open-source 5.9 GHz measurements with GPS coordinates are used for training.Source: 8) AI Enabled Railway Channel Modeling · p. 14
  • Spatial pyramid pooling maps variable-size image outputs to fixed-size feature vectors, while a 50-layer residual network with CBAM attention selects useful satellite-image features.The pooling design retains spatial information for subsequent fully connected layers.Source: 8) AI Enabled Railway Channel Modeling · pp. 14–15
  • 6.36 dB test-set RMSE, with a 2.5 dB relative improvement over a model trained using only satellite images of the area near the receiver.The complete-link transceiver satellite image is used as input.Source: 8) AI Enabled Railway Channel Modeling · p. 15
  • 2.2 dB relative RMSE improvement is reported under NLOS conditions for the model using the complete-link transceiver satellite image.Fig. 7 presents the NLOS comparison.Source: 8) AI Enabled Railway Channel Modeling · p. 15, Figure 7 · p. 15
  • Maximum street-level RMSE is at most 11 dB with complete-link information, versus 14.2 dB with receiver-area information, indicating better robustness.The per-street RMSE comparison is shown in Fig. 8.Source: 8) AI Enabled Railway Channel Modeling · p. 15, Figure 8 · p. 15

C. RIS for High-Mobility Scenarios

In high-mobility rail scenarios, channel aging reduces spectral efficiency, while RIS can mitigate that decline. The paper also surveys railway channel technologies and describes OTFS-TSMA as an approach to reliable massive access under limited spectrum.

C. RIS for High-Mobility Scenarios

  • Average SE falls over time, and increasing normalized Doppler shift accelerates the decline; adding RIS elements increases average SE and can alleviate channel aging.Sum SE also degrades with transmission time and normalized Doppler shift, with sharper declines at longer transmission times; RIS can alleviate the negative effect.Source: C. RIS for High-Mobility Scenarios · p. 16, Figure 11 · p. 16, Figure 12 · p. 16

D. Lessons Learned

  • Railway 6G channel technologies combine satellite, UAV, and THz connectivity with cross-domain modeling, RIS, AI tools, and IoT-R, while practical deployment challenges for THz and VLC remain unaddressed.The discussion identifies these technologies for connectivity, propagation issues, channel performance, predictive analysis, and real-time monitoring.Source: D. Lessons Learned · p. 16

A. OTFS in Smart High-Speed Railway

  • High-speed mobility and limited spectrum constrain railway umMTC, while OFDM is vulnerable to Doppler and OTFS is suited to significant delay and Doppler spread.OTFS-based multiple-access schemes still face complexity and realistic channel-characterization challenges.Source: A. OTFS in Smart High-Speed Railway · p. 17

1) Orthogonal time frequency space modulation enabled tandem spreading multiple access (OTFS-TSMA):

  • OTFS-TSMA targets reliable, low-complexity massive railway access by assigning users unique tandem spreading combinations and using erasure coding to recover limited collisions.Chip-level interleaving preserves spreading-sequence orthogonality under OTFS’s two-dimensional cyclic shifts; Doppler-domain diversity helps mitigate false alarms.Source: 1) Orthogonal time frequency space… · p. 17

2) Channel spreading function characterization in smart railways:

The section characterizes measured high-speed-railway channels and describes OTFS-TSMA and THz scheduling for reliable, efficient railway communications. It reports improved OTFS-TSMA BER performance over OFDM-TSMA and favorable scheduling comparisons against two baselines.

2) Channel spreading function characterization in smart railways:

  • Measured HSR channel spreading functions are less sparse and compact than TDL-model functions, with sampling error, SFFT effects, and small-scale fading contributing to the difference.The authors argue that discrete channel spreading properties in spectrum-limited systems should guide OTFS modulation design.Source: 2) Channel spreading function… · pp. 17–18
  • OTFS-TSMA outperforms OFDM-TSMA in HSR BER comparisons, with stable delay-Doppler channel gain mitigating deep-fading effects.The proposed tandem spreading combinations support access for 560 users; the passage also notes receiver design remains under development for reduced-sparsity, reduced-compactness channels.Source: 2) Channel spreading function… · p. 18, Figure 15 · p. 19

B. THz Communications for High-Speed Mobile Scenarios

  • THz links in vacuum-tube HSR can provide very high bandwidth, but propagation loss and blockage motivate robust transmission scheduling.The described system places remote access units outside the tube and uses steerable directional antennas on train-mounted MRs; 340 GHz is selected for the example system.Source: B. THz Communications for High-Speed… · pp. 18–20
  • The proposed location-aware scheduler orders RAU-relayed links to minimize scheduling time while satisfying each flow’s QoS requirement.It assigns link priorities and schedules flows in descending priority order; the scheduling problem is described as nonlinear mixed-integer and NP-hard.Source: B. THz Communications for High-Speed… · p. 20
  • The location-aware QoS scheduling scheme uses the fewest time slots and achieves the highest system throughput among the three evaluated schemes.The comparisons are against the QoS-aware concurrent scheme and MQIS; the figures plot slots against requested flows and throughput against flow count.Source: B. THz Communications for High-Speed… · p. 21, Figure 17 · p. 21, Figure 18 · p. 21

1) Semantic Communication in Railway Systems:

The section describes semantic-communication approaches for railway video, CSI feedback, and multimodal data fusion, alongside edge-intelligence architectures and methods. It presents these approaches as responses to bandwidth, channel-variation, resource, and deployment challenges in railway systems.

1) Semantic Communication in Railway Systems:

  • Rail-SC splits video intrusion detection between a computationally constrained railway device and an edge server, transmitting task-relevant semantics for downstream detection.A video Transformer models spatial and temporal correlations, while SNR-based layer normalization adapts intermediate features to channel conditions.Source: 1) Semantic Communication in Railway… · p. 21

2) Rail-SC System for MIMO with the CSI feedback:

  • For high-speed CSI feedback, ADJSCC-CSI jointly codes source and channel information and uses SNR adaptation to target low-overhead CSI reconstruction across channel conditions.The cited framework uses a single trained model with SNR information incorporated at transmitter and receiver.Source: 2) Rail-SC System for MIMO with the… · pp. 21–22
  • A hierarchy-aware, channel-adaptive semantic fusion framework dynamically combines fused-modal or deep and shallow features to reduce redundancy without additional bandwidth.It is intended for bandwidth-constrained railway tasks requiring modalities such as sensor readings and video.Source: 2) Rail-SC System for MIMO with the… · p. 22

D. Lessons Learned

  • The lessons-learned passage identifies OTFS, OTFS-TSMA, grant-free access, THz, semantic communication, channel modeling, and location-aware scheduling as technologies for railway transmission challenges.It also notes path loss, link blockages, and dynamic conditions as challenges requiring relay mechanisms and scheduling algorithms.Source: D. Lessons Learned · p. 22

V. INTELLIGENCE AND SECURITY TECHNOLOGIES FOR 6G RAILWAYS

  • The paper frames AI analytics and machine learning as supporting railway monitoring, prediction, and optimization, while encryption and intrusion detection protect system integrity and reliable operations.Source: §V · p. 22

A. Edge Intelligence Technologies

  • The section identifies varied railway channel settings, costly upgrades to linear base-station deployments, heterogeneous service requirements, and competition for wireless and MEC resources as edge-intelligence challenges.Source: A. Edge Intelligence Technologies · p. 23
  • The edge-intelligence architecture separates onboard mobile edge entities, which provide low-latency services and channel-access prediction, from edge data centers that complement unmet needs.The section notes that simultaneous task offloading by many terminals can cause interference and MEC queue congestion, increasing transmission errors and processing delays.Source: A. Edge Intelligence Technologies · p. 23

1) Single server application scenario:

The section develops federated-learning coalition approaches for single- and multiple-server railway scenarios, optimizing network utility alongside execution efficiency and reporting improvements over comparison algorithms.

1) Single server application scenario:

  • The single-server task-driven coalition algorithm jointly optimizes total network utility and efficiency for multi-task federated learning.It is especially applicable when trains stop at a station and the server has wide coverage.Source: 1) Single server application scenario · p. 24
  • Above 80% average accuracy was achieved, alongside reduced total task execution time and improved total network utility with the proposed single-server coalition algorithm.Fig. 24 shows convergence of the multi-task FL algorithms.Source: Figure 24 · p. 24, 1) Single server application scenario · p. 24

2) Multiple servers application scenario:

  • For multiple servers, block coordinate descent alternates bandwidth and computing-resource optimization, while stable matching forms many-to-one task-user coalitions.Task preferences account for training delay and local-parameter upload time.Source: 2) Multiple servers application scenario · p. 25
  • The proposed multi-server algorithm reaches a stable vehicular coalition and improves performance over random, K-means, and one-by-one execution algorithms.Fig. 25 depicts delay as the user coalition changes.Source: 2) Multiple servers application scenario · p. 25

3) Parameter-authentication federated learning:

The section presents parameter-authentication federated learning to address the tension between secure aggregation and privacy under malicious attacks. Its scheme uses zero-knowledge proofs and blockchain-based checks, and is reported to outperform existing schemes on several dimensions.

3) Parameter-authentication federated learning:

  • PAFL uses zkSNARK zero-knowledge proofs to authenticate federated-learning models while aiming to resist malicious attacks and protect client privacy.The single-message proof is presented as suitable for high mobility and strict delay requirements.Source: 3) Parameter-authentication federated… · p. 25
  • Users verify global-model genuineness, while the smart contract checks local-model proofs and drops invalid models; blockchain identities keep users anonymous.The scheme also authenticates local models against registered data, and invalid proofs affect vehicular reputation.Source: 3) Parameter-authentication federated… · p. 26
  • The scheme performs better than existing privacy-preserving and attack-defense federated-learning schemes in accuracy, attack defense, overhead, and privacy protection.Fig. 26 presents accuracy under defensive schemes.Source: 3) Parameter-authentication federated… · p. 26

B. Endogenous Security Technologies for Smart Railway

The section argues that future 6G smart railways need security built into the network architecture, rather than relying on patch-based protection. It proposes a coordinated endogenous-security framework and surveys supporting technologies.

B. Endogenous Security Technologies for Smart Railway

1) Endogenous secure railway network architecture:

  • The proposed cloud-edge-end architecture integrates security from the design phase, with AI monitoring risks and coordinating proactive detection and intrusion prevention.The framework combines trusted-computing hardware, blockchain-powered data consensus, IoT security awareness, and AI-based attack defense.Source: Figure 27 · p. 27, 1) Endogenous secure railway network… · p. 27

2) Key technologies for endogenous security in a smart railway:

  • The section identifies intelligent analysis, trusted authentication, and distributed collaboration as technologies for end-to-end railway security across space-ground networks.The stated capabilities include auditing, detection, response, and recovery.Source: 2) Key technologies for endogenous… · p. 27
  • Examples include AI-based abnormal-traffic detection, verified-device access, and blockchain-supported distributed caching and trust management.The cited applications also include decentralized access control and distributed reinforcement learning for trusted services.Source: 2) Key technologies for endogenous… · pp. 27–28

3) Blockchain empowered trustworthy railway networks:

The section examines blockchain-based security and digital-twin (DT) networks as ways to address railway security and operational challenges. It outlines DT architectures and learning schemes while identifying modeling, synchronization, and security constraints.

3) Blockchain empowered trustworthy railway networks:

  • Conventional railway designs may leave identity, access, communication, and data transmission vulnerable when security is not considered from the outset.The section describes a blockchain-enabled, zero-trust architecture with distributed identity storage, registration, two-way authentication, reputation evaluation, and zero-knowledge proofs.Source: 3) Blockchain empowered trustworthy… · p. 28

C. Railway Digital Twin Networks

  • DT networks can support railway monitoring and optimization, but comprehensive modeling remains difficult with many devices, dynamic systems, and limited resources.The section also identifies possible divergence between DTs and physical entities due to losses during data transmission and training.Source: C. Railway Digital Twin Networks · pp. 28–29
  • Railway DT implementations also face security and privacy risks from sensitive data transmissions and the high-speed, time-varying railway environment.The section notes that Doppler shift and changing network topology and communication links complicate implementation; geographically distributed, correlated sensor data raise controllability and security challenges.Source: C. Railway Digital Twin Networks · pp. 29–30

1) Digital twin system modeling:

  • The railway DT architecture links physical trains, sensors, and terminals with virtual twins on edge and cloud servers for real-time synchronization and feedback.MEC networks are expected to provide the low latency, reliability, and security needed for this interaction; associating DTs with servers and allocating communication resources remain challenges.Source: 1) Digital twin system modeling · p. 30
  • Federated learning trains DT models collaboratively, while a permissioned blockchain records DT data and controls client participation.BSs train local models and send parameters to an MBS; BSs maintain the blockchain, and the MBS serves as the federated-learning server.Source: 1) Digital twin system modeling · p. 30

2) Blockchain empowered DT:

  • The blockchain-enabled DT framework records DT models, DT data, and training models, with peer BSs verifying local models before global aggregation.Training models are shared as blockchain transactions; DPoS consensus and performance-linked training coins govern transaction validation and block production.Source: 2) Blockchain empowered DT · pp. 30–31

3) DT edge association:

The section frames DT edge association as a resource-allocation problem for synchronizing railway clients with DTs hosted at BSs. It describes the factors shaping association and the learning workflow, then notes potential communication benefits.

3) DT edge association:

  • Edge association maps railway devices to BSs according to dataset, BS computing capacity, and the client-to-BS transmission rate.The association is intended to balance learning accuracy against learning time while improving DT-system efficiency and accuracy.Source: 3) DT edge association · p. 31
  • DT construction involves local training, data transmission, and model aggregation, with training time depending on BS computing capability and DT data size.BSs aggregate local models and broadcast them as transactions; the described workflow also includes block validation by elected producer BSs.Source: Figure 33 · p. 31, 3) DT edge association · p. 31
  • The frameworks are described as enabling decentralized, privacy-preserving channel-estimation training and DT-based simulation of dynamic HSR network environments.The section says these simulations facilitate more accurate and adaptive channel-estimation methods.Source: 3) DT edge association · pp. 31–32

E. Lessons Learned

For smart-railway access, the section examines AI-based methods for changing user activity and data lengths, reporting improved channel estimation and data recovery in the described comparisons. It concludes that synchronization, spectrum, and security challenges remain.

E. Lessons Learned

  • The section identifies synchronization errors, limited spectrum, and data security as continuing challenges for IoT in next-generation railways.Its discussion connects mMTC with monitoring and predictive maintenance, and describes sporadic activity, small packets, and high device density as relevant characteristics.Source: A. Massive Machine Communication for… · p. 32, E. Lessons Learned · p. 32

1) Dynamic changes in the number of active users:

  • Passing trains create a bimodal distribution in active-user numbers, which the described LAMP approach uses to learn data structure for detection and channel estimation.The access model assumes contention-based pilot selection, synchronized packets, and block-fading channels within each coherence time.Source: 1) Dynamic changes in the number of… · p. 33

2) Data length diversity:

  • Different packet lengths create varying sparsity across data columns, which BSAMP-CP and modified LAMP exploit for joint detection, channel estimation, and data recovery.The described backward approach estimates sparsity before repeatedly using pilot and data joint sparsity, error-checking information, and modulation-constellation information.Source: 2) Data length diversity · p. 33
  • LAMP has advantages over OMP, ISTA, and AMP for channel estimation, and does not require many iterations after training.The passage reports that all methods improve as SNR increases, while ISTA gradually stabilizes; it describes LAMP as more suitable when active-user numbers change dynamically.Source: Figure 35 · p. 33, 2) Data length diversity · p. 34
  • At 30 dB SNR, BSAMP-CP’s data recovery ratio can reach twice that of SAMP.The comparison considers SP, CoSaMP, SAMP, and BSAMP-CP for data recovery across varying SNRs.Source: 2) Data length diversity · p. 34, Figure 36 · p. 34

B. Integrated Sensing and Communications

The section presents integrated sensing and communications as part of a broader 6G railway network combining sensing, communication, computing, storage, and intelligence. It discusses integration challenges, resource coordination, security, and cell-free architectures.

B. Integrated Sensing and Communications

  • Railway ISAC combines communication and sensing, with channel estimation supporting beamforming, positioning, tracking, obstacle detection, and infrastructure monitoring.The section describes potential benefits including reliable communication and large-scale sensing alongside reduced hardware costs, spectrum use, and energy consumption.Source: B. Integrated Sensing and Communications · p. 34, Figure 37 · p. 34
  • Railway SCCSI design still faces three key challenges, including the need to integrate communication, sensing, computing, storage, and intelligence.The section describes high-speed rail safety as requiring both reliable low-latency communications and comprehensive, high-precision sensing.Source: B. Integrated Sensing and Communications · p. 34

1) Integrated Design of Sensing, Communication, Computing, and Intelligence for HSR Wireless Communications:

3) Joint Optimization of Sensing, Communication and Computing Resources for Intelligent Railway:

  • Perception research often emphasizes communication and computing resource optimization more than the perception process, although tasks can couple, collaborate, or compete across all three functions.The cited examples address joint communication-sensing-computation allocation and combining over-the-air computing with ISAC.Source: 3) Joint Optimization of Sensing,… · p. 35

4) Security and Privacy for the Integration of Intelligent Railway Communication, Sensing and Computing:

  • Railway sensing and communication infrastructure faces threats to availability, integrity, and confidentiality, and integrated designs may facilitate eavesdropping.The section proposes low-complexity data protection using privacy technologies including zero-knowledge proofs, federated learning, and secure multi-party computation.Source: 4) Security and Privacy for the… · p. 35

A. Cell-Free Network Architectures

  • Cell-free massive MIMO is presented as suited to high-speed rail, but a fast-switching solution remains necessary for ultra-reliable, low-latency communication.Distributed massive MIMO can improve coverage and reduce base-station switching, while the cited static-scenario studies overlook subcarrier interference.Source: A. Cell-Free Network Architectures · p. 36

B. Channel Modeling and Characteristics for All Frequency Bands and All Scenarios

The article reviews 6G railway requirements and technologies, proposing an integrated architecture and identifying open research directions. It surveys enabling methods across channel modeling, security, edge intelligence, URLLC, digital twins, and generative AI.

B. Channel Modeling and Characteristics for All Frequency Bands and All Scenarios

  • Cross-domain railway channel modeling must span frequencies and scenarios despite sparse dynamic data, with AI supporting real-time clustering, recognition, prediction, and modeling.Source: B. Channel Modeling and… · p. 36

C. Endogenous Security for 6G Smart Railways

  • 6G endogenous security shifts from Internet to cyberspace security, embedding AI monitoring and threat prediction, while blockchain, private computing, and distributed collaboration require integration and resource-allocation study.Source: C. Endogenous Security for 6G Smart… · p. 36

D. Connected Cooperative Edge Intelligence

  • Connected cooperative edge intelligence faces limited computation and communication resources, motivating parallel, resource-efficient learning and improved allocation strategies.Source: D. Connected Cooperative Edge… · p. 36

E. New URLLC Techniques

  • Super-high-speed trains exceeding 1000 kilometers per hour require new URLLC frame structures under limited block code length to balance ultra-reliability and ultra-low latency.Source: E. New URLLC Techniques · pp. 36–37

F. MEC Empowered Digital Twin Networks

  • MEC-assisted railway digital twins face challenges in near-real-time model construction with limited resources, simulation error, and low-latency model placement and transfer under high mobility.Source: F. MEC Empowered Digital Twin Networks · p. 37

G. Generative AI Applications

  • Generative AI may support railway channel modeling, data augmentation, anomaly detection, and lightweight edge models, but computational complexity, privacy, and high-quality data remain challenges.Source: G. Generative AI Applications · p. 37

VIII. CONCLUSION

  • The article proposes an integrated network architecture spanning wireless coverage, storage, computing, and communication, and surveys enabling technologies, deployment methods, and open research issues.Source: §VIII · p. 37
Loading 2505.12946v1…