network simulation

Traditional 3GPP statistical channel models lack scene-specific fidelity—e.g., street canyons, corner diffraction, or deterministic line-of-sight (LOS) propagation—leading to insufficient accuracy in system-level simulations. To address this, we propose a trajectory-driven, site-specific channel modeling framework that tightly couples the Sionna ray tracer with ns-3/5G-LENA. This integration converts geometrically accurate multipath components into frequency-domain channel matrices and embeds them seamlessly into the PHY/MAC protocol stack, preserving full 3GPP standard compliance while enabling high-fidelity simulation. The method supports digital twin research from 5G-Advanced to 6G. Experimental validation demonstrates its superior capability in capturing beamforming performance and end-to-end key performance indicators—including throughput knee points and coverage outages—significantly outperforming conventional statistical models.

3GPP TR 38.901 exhibits insufficient modeling capability in the 7–24 GHz band, limiting its applicability to 6G system design. Method: Targeting 3GPP Release-19 standardization, this work systematically enhances TR 38.901 by introducing cluster/ray dynamic variability, terminal antenna pattern coupling, multi-polarized power distribution, near-field propagation effects, and spatial non-stationarity modeling—within a geometric stochastic framework and validated against extensive measurement data, thereby relaxing conventional far-field and wide-sense stationarity assumptions. Contribution/Results: The enhanced model significantly improves physical-layer simulation accuracy in representative scenarios such as suburban macrocells, bridges the standardization gap in channel modeling across the Sub-6 GHz to millimeter-wave transition band, and establishes an authoritative, scalable foundation for link-level and system-level performance evaluation in 6G.

Existing network simulators (e.g., ns-3) suffer from physical inaccuracies in wireless channel modeling—particularly in complex indoor/outdoor environments—failing to faithfully capture diffraction, multipath propagation, and spatiotemporal correlation. Method: This work presents the first end-to-end, GPU-accelerated integration of the Sionna RT ray-tracing engine with ns-3, enabling physics-based 3D channel modeling grounded in scene geometry and device positions. We propose a channel coherence-time–driven intelligent pre-caching mechanism to drastically reduce real-time ray-tracing overhead, and support fine-grained CSI output and ray-tracing–informed mobility modeling. Results: Experiments demonstrate >40% reduction in path-loss and multipath delay estimation errors compared to conventional ns-3 models; spatiotemporal correlation modeling accuracy is substantially improved. The framework efficiently supports integrated sensing and communication (ISAC) and medium-scale mobile network simulation.

Existing 5G system-level simulators struggle to balance simulation speed and seamless integration with AI frameworks, hindering co-design research at the intersection of wireless communications and machine learning. To address this, we propose a novel computation-graph-based simulation paradigm that replaces traditional discrete-event simulation with a directed graph composed of differentiable computational blocks, enabling demand-driven, intelligent update strategies—thereby significantly improving both efficiency and flexibility. Implemented entirely in pure Python with a modular architecture, the simulator natively supports mainstream AI frameworks such as PyTorch and TensorFlow. Our open-source simulator preserves physical modeling fidelity while accelerating typical scenario simulations by one to two orders of magnitude. This substantially lowers the technical barrier for joint development of communication algorithms and deep learning models, providing an efficient, scalable, and AI-native system-level simulation infrastructure for 6G intelligent wireless networks.

This work addresses the limitations of the 3GPP-standardized tapped delay line (TDL) channel model in accurately capturing the spatial propagation characteristics of MIMO systems, which can lead to biased performance evaluations. To overcome this, the authors propose and validate a reduced cluster delay line (rCDL) model. Through comparative analysis against real-world channel measurements in representative commercial scenarios, the study evaluates the spatial modeling accuracy of rCDL relative to TDL and further assesses their discriminative capability via CSI reporting performance simulations. Results demonstrate that rCDL significantly improves the fidelity of spatial channel characterization while maintaining reasonable computational complexity. It outperforms TDL in both measurement-to-model alignment and evaluation of CSI feedback schemes, thereby offering strong support for future 3GPP standardization efforts.

研究通过NS3Learn模型解决了5G NR Mode-2在车联网安全评估中不考虑资源竞争导致的高估消息传递成功率问题，提高了密集交通下的通信仿真真实性。

This study addresses the ambiguous standardization pathway and technical challenges facing 6G Integrated Sensing and Communication (ISAC) by systematically reviewing 3GPP Release 19–20 evolution, encompassing requirements, channel models, and protocol architectures. By pioneering the alignment of service requirements with physical layer and protocol design, and integrating RAN1/2/3 investigations with sensing-assisted communication analysis, this work establishes a standard-oriented 6G ISAC technical framework. The research identifies critical unresolved issues in current standardization efforts and proposes a feasible evolutionary roadmap. Ultimately, these findings provide essential theoretical underpinnings and practical guidance for advancing 6G ISAC standardization, bridging the gap between high-level service demands and concrete technical specifications within the evolving 3GPP landscape.

This study addresses the ambiguity in translating 6G visions into concrete specifications and the lack of industry consensus during 6G standardization. By systematically analyzing early 3GPP 6G initiatives, this work examines the normative evolution across the RAN, SA, and CT working groups concerning key modules such as service requirements, native AI, integrated sensing and communication (ISAC), and non-terrestrial networks (NTN). The contribution lies in distilling seven cross-domain insights that establish AI as a foundational primitive, NTN as an underlying architecture, and the convergence of communication, computing, and sensing as a paradigm shift. Furthermore, it delineates a Release-20/21 roadmap targeting commercial deployment by 2030, advocating interoperability testing as a filtering criterion to substantively advance 6G from the research phase toward engineering specification.

This study addresses the lack of scalable multi-hop 5G NR sidelink mesh network simulation capabilities in OpenAirInterface (OAI) by proposing a component-based SL-RFSIM framework. This framework replaces conventional RF simulators with a broker-based publish/subscribe architecture, supporting arbitrary point-to-point connections and Layer-2 mesh networking. Through a pluggable design integrating modular services such as mobility and propagation models, it establishes a large-scale multi-hop experimental environment fully compatible with the OAI protocol stack. Validation using the BATMAN-adv routing protocol on the SLICES-RI infrastructure demonstrates that the proposed architecture effectively supports large-scale sidelink simulations while identifying critical software bottlenecks. The project has been released as open source, providing a reproducible experimental foundation for future research.

研究通过调整TNGF将Wi-Fi整合进5G网络切片，使用开源5G核心网和非3GPP接入进行验证，解决了多RAT环境下的统一服务提供问题。