Low Earth orbiting (LEO) satellite-assisted wireless communication is increasingly vital for future communication networks due to the significant spectrum scarcity in radio frequency channels, presenting a critical bottleneck. Thus, optimizing the utilization of available radio frequency spectrum has become imperative. Advanced techniques like underlay communication and Rate Split Multiple Access (RSMA) have proven effective in enhancing spectrum utilization. When LEO satellites are applied to tasks such as agricultural assistance, search and rescue operations, and military defense, LEO-to-ground communication can leverage underlay fashion using RSMA to transmit messages to multiple users simultaneously on the same channel. However, conventional underlay communication setups necessitate transmitter cooperation to manage system interference. Enabling non-cooperative systems to communicate in an underlay fashion unlocks the untapped potential of these advanced transmission techniques. This study addresses the challenge of maximizing the RSMA rate of the LEO-to-ground communication system (secondary system) operating in an underlay mode without cooperation with the ground-to-ground communication system (primary system), where the primary network operates in a time-division multiple-access fashion. We propose a dueling-based double deep Q-learning solution to optimize the allowed transmission power at the LEO satellite, ensuring no outage in the primary system. Additionally, we introduce an optimal solution framework to distribute the allowed transmission power among all signals of the secondary devices, maximizing the RSMA rate while meeting the rate requirements of all underlay secondary devices. Simulation results demonstrate that this hybrid solution framework provides excellent performance while ensuring no outage at the primary network.

Deep neural networks (DNNs) have been widely used in recent years for wireless communication applications, including channel estimation. However, DNN performs well for environments (data) that it has been trained for. However, it is still challenging to work in a non-stationary system where the statistical characteristics of the environment change with time. This paper introduces an online adaptation approach that allows the neural network to dynamically change to cope with the non-stationary channel estimation. The idea is to use the existing detected packets to continuously update the weights of the DNN. Simulation results show that our approach can significantly improve the performance of the DNN in non-stationary environments whereas concurrent state-of-the-art algorithms fail.

In this paper, we propose a deep learning autoencoder (AE) to model and design space optical communications (SOC) systems from end-to-end performance. The proposed AE is based on multiple-decoders and a new layered structure for constructing both encoders and decoders. The use of multiple-decoders can increase the receiver diversity, which allows the gradient descent to minimize the cost function compared to state-of-the-art models and therefore improve the bit error rate (BER) performance. In addition, we utilize system tool kit (STK) simulator for a realistic SOC channel modeling in Log-normal fading and additive white Gaussian channel. Numerical results reveal that the traditional convolution codes-based SOC systems are outperformed by the proposed learning-based AE for various code rates. Moreover, the proposed AE outperforms the benchmark AE learning-based models in terms of BER performance metric.