The growing number of remote sensing (RS) satellites and their enhanced sensing capability lead to exponential growth in RS data, posing challenges to real-time transmission. To address this issue, this paper establishes a three-layer space-earth collaborative architecture comprising RS satellites, communication satellites, and earth stations. Specifically, we propose a joint optimization framework that jointly optimizes compression strategy selection and computing resource allocation, with the aim of minimizing long-term end-to-end delay. However, the optimization problem is a mixed integer non-convex problem with coupled variables, making it difficult to obtain the global optimal solution. To solve the problem, we firstly apply Lyapunov optimization theory to decompose the long-term problem into the deterministic sub-problem in each slot. Subsequently, the compression selection variables are decoupled from the resource allocation variables, and the resource allocation variables are solved via convex optimization to obtain a dimensionality-reduced Markov decision process (MDP) problem. An intensive reward reinforcement learning algorithm is proposed to address the reward sparsity problem in the MDP. Finally, simulation results demonstrate that the proposed algorithm significantly improves performance and reduces the average delay compared to the benchmarks.

The digital transformation of industrial systems has been significantly influenced by the emergence of the Industrial Internet of Things. Digital twin (DT) technology plays a pivotal role in the transformation, serving as a bridge between the physical and digital realms. To support the efficient application of DT technology, the synchronization between physical entities (PEs) and DT models (DTMs) can not be ignored. Given the open nature of wireless channels, the diversity in synchronization mechanisms arising from the functional variations among PEs inevitably deteriorates the design of synchronization strategies. Moreover, a metric is required to gauge the synchronization between PEs and DTMs. In this article, a reinforcement learning (RL)-based online scheduling scheme is proposed to achieve efficient synchronization between DTMs and PEs with different mechanisms. Specifically, the Age of Information (AoI) is introduced as a metric to evaluate synchronization strategies. In addition, PEs are classified and share spectrum resources according to synchronous mechanisms, improving resource utilization efficiency. To ensure real-time scheduling, we propose a hybrid inverse RL-based scheme to support distributed time slot-level synchronization, reducing the need for manual intervention. Simulation results show that compared with other baseline RL schemes, the proposed scheme can reduce the AoI value more than 20% between the PE and the DTM.

The Non-geostationary orbit (NGSO) satellite constellation has gained widespread recognition owing to its exceptional low latency and seamless global coverage. However, with the increasing number of countries launching satellites, there has been a surge in the amount of satellites orbiting in low earth orbit, resulting in a growing strain on both frequency and orbital resources. Therefore, the issue of satellite coexistence has become increasingly critical. Moreover, the uneven distribution of terrestrial users imposes higher demands on beam resource management. This paper proposes a beam allocation strategy based on matching theory, which effectively mitigates harmful interference from the main constellation while optimizing the throughput of the minor constellation. The proposed strategy enables on-demand allocation, which maximizes both service quality and resource utilization, ultimately achieving the objective of constellation coexistence. Simulation results demonstrate that the performance of the proposed strategy is significantly superior compared to other link establishment schemes.

Federated learning (FL) is a paradigm that enables model training across various devices while keeping the data localized. However, for battery-powered passive devices in the satellite Internet of Things (IoT), the continuous update and transmission of the local model result in heightened energy consumption on the device side. To address this challenge, an FL framework with partial device participating is proposed. In this framework, the on-board controller strategically selects a subset of devices to upload local model parameters, effectively mitigating the overall energy consumption on the device side. Constrained by transmission power and transmission delay, a resource allocation problem is formulated. This problem jointly optimizes the uploading strategy and transmission power, aiming to minimize the utility function that combines the global model loss and energy consumption over multiple rounds of FL. Simulation results demonstrate that, compared with other benchmark schemes (DDPG, PPO), the proposed algorithm achieves energy efficiency in FL.