This paper presents a general framework for probabilistic relay selection (RS) in asymmetric buffer-aided cooperative relaying systems, which caters to scenarios with both perfect and imperfect channel state information (CSI) during the RS process. The framework extends and generalizes many existing buffer-aided RS schemes. In particular, we introduce an auxiliary stochastic process which assigns varying selection probabilities to different links, considering the dynamic wireless channel and buffer states. Subsequently, we leverage the obtained outage probability and average packet delay (APD) to formulate outage optimization problems while adhering to APD. To address the intricate high-dimensional optimization problems, we employ a deep learning (DL) approach, which involves designing probability mass functions for the auxiliary stochastic process and developing an effective loss function to update the neural network. Simulation results unequivocally demonstrate the superior performance of the proposed DL-based probabilistic RS scheme compared to benchmark schemes, particularly in scenarios involving imperfect CSI.

In this paper, we study the simultaneous wireless information and power transfer (SWIPT) cooperative system, where one source forwards information to one destination with the assistance of multiple relays. Each relay is equipped with a finite data butter and a finite energy butter storing the harvested energy by radio-frequency (RF). An optimization problem is formulated for throughput maximization of the SWIPT cooperative system, taking into consideration the strict delay constraint, dynamic channel conditions, time-varying discrete data butter states and time-varying continuous energy butter states. A discrete-time Markov decision process (MDP) is adopted to model the relay selection process referring to data butter states and energy butter states. Two deep Q-network (DQN)based methods named invalid action penalty (IAP) and invalid action mask (IAM) are proposed. The simulation results show that the proposed IAM method can achieve better convergence and throughput performance than the IAP method.