In large-scale Low Earth Orbit (LEO) satellite networks, cache placement and update are facing challenges such as limited cache volume, update capacity, frequently interrupted Inter-Satellite Links (ISLs) and sudden changes in content popularity, leading to the difficulty for obtaining fresh data for existing algorithms by optimizing average user access delay instead of timeliness and adaptability. To address this issue, in this work, we propose an age-driven Multi-Agent Deep Reinforcement Learning (MADRL) based cooperative cache and update mechanism called as Age-Driven Cooperative Cache and Update (ADCCU) algorithm, in which we establish an age-driven cache gain function to evaluate effective value of holding a data item from cache’s perspective. In particular, we formulate the age-driven cooperatively cache scheduling issue as an Integer Programming (IP) problem and solve it by exploiting a Markov Decision Process (MDP). Simulation results demonstrate that, under a cache capacity constraint ofci=3, the ADCCU algorithm significantly outperforms baseline caching strategies: the average cache gain is increased by approximately 1.09 compared to the Deep Deterministic Policy Gradient (DDPG) based cache, 2.88 compared to the Least Frequently Used (LFU), and 4.92 compared to the Least Recently Used (LRU), respectively, and the cache hit ratio is improved by approximately 5.31% over the DDPG-based cache, 56.4% over LFU, and 65.9% over LRU, respectively.

In the upcoming global-coverage 6G networks, high packet loss and long latency in long-distance transmissions exacerbate the trade-off between data timeliness and integrity, particularly in time-sensitive applications involving time-series data with stringent integrity requirements. This challenge exposes the limitations of existing transmission systems, such as source-channel coding and semantic communication, which fail to jointly address both dimensions. In this paper, we propose a deep learning (DL)-enabled generalized deduplication (GD)-based semantic transmission (DLGD-ST) mechanism for time-series data. By leveraging GD to address the impact of semantic ambiguity on data integrity, DLGD-ST exploits the semantic recovery and temporal discreteness of the data to effectively mitigate the conflict between integrity and timeliness. In particular, a well-designed long-short-term memory (LSTM)-based GD algorithm is developed to separate shallow semantic components and supplementary components, ensuring the integrity of semantic transmission. A deep semantic encoding process is then performed using a double-layer progressive dimension reduction (DPDR) and adaptive quantization (AQ) scheme, which capitalizes on the channel robustness of semantics to reduce transmission rounds and improve timeliness. Furthermore, an incremental dimension hybrid automatic repeat request (ID-HARQ) mechanism is introduced to improve semantic reliability by retransmitting high-dimensional semantics, thereby further minimizing end-to-end transmission rounds. To accurately evaluate performance, we introduce the Age of Semantic Information (AoSI), which incorporates integrity constraints into the generalized Age of Information (AoI) to jointly assess integrity and timeliness. Simulation results demonstrate that the proposed DLGD-ST mechanism, enabled by accurate data recovery and reduced transmission rounds, achieves better AoSI performance compared to existing communication systems under both high and low signal-to-noise ratio (SNR) conditions.

In the Large-scale Satellite Network (LSN), inter-beam interference significantly hinders the transmission performance of Low Earth Orbit (LEO) satellite downlinks. This interference exhibits notable time-varying characteristics due to the relatively rapid movements between LEO satellites and ground users, posing a substantial challenge to existing transmission resource allocation techniques. To tackle this issue, we introduce the Hypergraph Neural Network (HGNN)-enabled Resource Allocation (HGNNRA) algorithm for the downlinks of LSN. This algorithm aims to solve a transmission-rate maximization problem, effectively handling the time-varying interference among multiple beams in the downlink through a well-tailored Dynamic Hypergraph Neural Network (DynHGNN). Specifically, considering the coupling complexity of satellite beam coverage and user participation, we have developed a Dynamic Hypergraph-based interference model, along with a customized construction algorithm, to describe their time-varying relationships precisely. Simulation results indicate that our proposed HGNNRA outperforms both Graph Convolutional Network (GCN) [14], HGNN [15], GNN-DDQN [48], and GCNRA in terms of transmission rate and the satisfaction degree of user transmission requirement metrics.

In the Large-Scale Heterogeneous Satellite Networks (LSHSNs) integrating Low Earth Orbit (LEO) and Medium Earth Orbit (MEO) satellites, data delivery faces complex topology and dynamic connectivity, which poses a significant challenge to current graph-dependable transmission strategies requiring global topological information, incurring huge computational cost and interactive overhead. To address this issue, in this paper, we propose a Local Information-dependable Semantic Delivery Mechanism (LISDM) by exploiting topological features at the semantic level, in which we develop a Task-oriented Semantic-aware Topology Compression Network (TSTCN) to condense the global topology according to specific task demands. Besides, we develop a Deep Q-Network enabled Semantic Coded Routing (DQNSCR) algorithm for the semantic delivery in the LISDM by designing a novel metric called Semantic Delivery Efficiency (SDE). The simulation results indicate that the proposed mechanism performs better in improving the required topology scale and throughput compared with typical data delivery mechanisms such as the conventional Open Shortest Path First (OSPF) routing algorithm, the DQN-based Intelligent Routing (DQN-IR) algorithm, and Real-Time Hop-by-Hop Routing (RTHop) algorithm with Space-Time Graph (STG) model, respectively.

In Large-Scale Heterogeneous Satellite Networks (LSHSNs) integrating Low Earth Orbit (LEO) and Medium Earth Orbit (MEO) satellites, high-timeliness data delivery confronts dynamical connectivity and obvious latency, which heavily challenges existing graph-dependable transmission strategies requiring to obtain global topological information with huge computational cost and signaling overhead. To address this issue, in this paper, we propose an Age-predicting Local Information Dependable Transmission (ALIDT) mechanism for the LSHSN by considering the impact of time-varying topology on the timeliness of data, in which a novel metric of data freshness called Forwarding-aware Age of Information (FAoI) is well-designed to evaluate the timeliness in data forwarding at node. In particular, we develop a satellite Coverage-based Local Information Sharing (CLIS)-assisted Spatial-Temporal Graph Neural Network (STGNN) to extract the topological features in both temporal and spatial dimensions and a Graph Matching Network (GMN)-based topology compression algorithm to improve computation efficiency. The simulation results indicate that the proposed mechanism performs better in improving the storage overhead, throughput and average FAoI compared with the conventional Open Shortest Path First (OSPF) routing algorithm with Time-Varying Graph (TVG) model, GNN-based Multipath Routing (GMR) algorithm, and Gated Recurrent Units (GRU) based metric prediction algorithm in hybrid satellite networks, respectively.

In long-span space communication, the current Licklider Transmission Protocol (LTP) confronts apparent challenges such as high packet loss rate and huge latency when carrying the bundles in the Disruption Tolerant Networks (DTN). These challenges incur obviously low freshness of satellite telemetry and instruction data with high timeliness requirements since the typical Automatic Repeat reQuest (ARQ) mechanism is exploited in the LTP for reliable transfer. To address this issue, in this paper, we propose an age-driven bundle delivery mechanism called as Semantic LTP (S-LTP) by considering the semantic correlations in the context-dependent data, which has excellent error-tolerant capability by a well-designed Semantic Supplement Hybrid Automatic Repeat reQuest (SS-HARQ), making it with high timeliness. In particular, a novel metric of semantic freshness of data called Age of Semantic Information (AoSI) is proposed to evaluate the timeliness contribution of information at the semantic level. The simulation results indicate that the proposed SS-HARQ scheme performs better in reducing the average AoSI and AoI by 62.24% and 64.52% respectively compared to the conventional LTP-ARQ with Cyclic Redundancy Check (CRC), 6.39% and 27.09% respectively compared to the Semantic Coding HARQ (SCHARQ) with a similarity detection network called Sim32.

In large-span space communication, the current bundle delivery mechanism using the Disruption-Tolerant Networks (DTN) technique confronts huge challenges such as high packet loss rate and huge latency. These challenges incur obviously low goodput when delivering scientific and engineering data for the target missions, such as instructions, text, and images. However, the context correlations in these data are not yet excavated to resist the above challenges by current works. To address this issue, therefore, we propose a semantic-aware bundle delivery mechanism for context-dependent data via a cross-layer design on Bundle Protocol (BP) and Licklider Transmission Protocol (LTP), which has the excellent error-tolerance capability by the well-designed semantic-oriented Automatic Repeat reQuest (ARQ) scheme. In particular, the jointed cross-layer design consists of a Semantic Blocking (SB) and Semantic Coding (SC)-based Bundle Updating (BU) mechanism and a dynamic Red/Green-part Allocation method based on Semantic Importance (RGA-SI) for bundles and segments in the two layers. Simulation results show that the proposed mechanism can reduce data latency and improve goodput from about 50% to 70% compared with the current bundle delivery mechanism in DTN with optimal segment size, especially under bad channel conditions.

In large-scale satellite networks, data delivery confronts obvious challenges such as high loss rate and long propagation delay leading to low Packet Delivery Ratio (PDR) and huge delivery latency over intermittent Inter-Satellite Links (ISLs), making the current routing algorithms exploiting typical Automatic Repeat reQuest (ARQ) mechanisms extremely inefficient and even incapable. To address this issue, in this paper, we propose a semantic-aware coding and routing joint mechanism called Semantic Adaptive Coding and Routing (SACR) by considering both the semantic correlations in the context-dependent data and the link status knowledge. In particular, the proposed SACR achieves excellent error-tolerant and routing-agile capabilities by an elaborately interactive design consisting of a customized routing-aware Semantic Adaptive Coding Hybrid ARQ (SAC-HARQ) mechanism and a Semantic Coding-based Routing Mechanism (SCRM). The simulation results indicate that the proposed SACR mechanism performs better in reducing the average delivery latency and improving the effective throughput compared with typical routing mechanisms such as Open Shortest Path First (OSPF) routing, Deep Q-Networks based Intelligent Routing (DQN-IR), and Real-Time Hop-by-hop routing (RTHop), integrating with typical semantic coding methods, i.e., Deep Learning-based Joint Channel-Source Coding (DL-JSCC), Deep learning-based Semantic Communication system (DeepSC), and Semantic Coding HARQ (SCHARQ), respectively.