In this paper, we consider real-time beamforming design for dynamic wireless environments with different channel state information (CSI) distributions in cell-free systems. Specifically, a sum-rate maximization optimization problem for different CSI distributions is built to model the beamforming design of dynamic wireless environments in cell-free systems. To efficiently solve the optimization problem, we propose a high-generalization network (HGNet). By preserving invariant features and discarding sensitive features for different CSI distributions, HGNet effectively improves the generalization performance of beamforming design for dynamic wireless environments in cell-free systems. Numerical results demonstrate that HGNet achieves a higher sum rate with a lower reflection time for different CSI distributions, thus realizing real-time beamforming design for dynamic wireless environments in cell-free systems.

Network slicing is a key enabler for 6G mobile networks. Guaranteeing the service level agreement with the smallest amount of radio resources is a challenging problem in network slicing scenarios due to random traffic patterns and the channel environment. To this end, we propose a novel deep reinforcement learning algorithm named CGDSAC based on the conditional generative diffusion model to achieve the optimization objective while capturing the underlying environment distribution. Subsequently, we further design a digital twin (DT) enhanced version of CGDSAC named CGDSAC-DT, to address issues that CGDSAC is unsafe or has lower performance than the default strategy in the early training stages, and converges slowly. Numerical results show that our proposed method can solve the issues encountered and outperform the baseline algorithm regarding performance metrics.

Network slicing is essential for the sixth-generation mobile networks. Minimizing radio resources while guaranteeing service level agreement (SLA) remains challenging due to random traffic patterns and channel conditions, making policy security enforcement and underlying traffic distribution inference critical research goals. In this paper, to facilitate the design of policy agents for radio resource management, we first design a high-fidelity conditional generative diffusion model (CGDM)-driven digital twin network (DTN) to provide closed-loop interaction and pre-validation capabilities. The DTN consists of a safety-bound coarse correction method to enhance strategy SLA compliance, a model market for agent warm-up and decision-level pre-validation, and a virtual interaction environment for high-fidelity agent pre-optimization. Then, a CGDM-driven safe reinforcement learning agent based on constrained multi-agent Markov decision process, termed CGD safe actor-critic (CGDSAC), is proposed to manage inter-slice radio resources. CGDSAC balances safety and strategy quality via Lagrangian primal-dual optimization and a behavior cloning objective targeting DTN-corrected strategies, while capturing latent traffic patterns. Furthermore, CGDSAC comprehensively leverages policy warm-up, decision-level pre-validation, and policy-level pre-optimization capabilities of DTN to resolve early inferior performance, SLA jitter, and slow convergence. Numerical results confirm that the built DTN exhibits good fidelity. Under fixed slices and stable traffic pattern, the DTN-enhanced approaches outperform the best baseline with an average SLA violation relative reduction of 71.3% and an average resource block utilization relative degradation of 10.9%, achieve about convergence speed enhancement of 80% compared to native CGDSAC, and is capable of adapting to the scenarios of dynamic number of slices and varying traffic patterns through knowledge transfer and pre-validation.

In this paper, we consider the problem of joint beam selection and link activation across a set of communication pairs to effectively control the interference between communication pairs via inactivating part communication pairs in ultra-dense device-to-device (D2D) mmWave communication networks. The resulting optimization problem is formulated as an integer programming problem that is nonconvex and NP-hard. Consequently, the global optimal solution, even the local optimal solution, cannot be generally obtained. To overcome this challenge, this paper resorts to design a deep learning architecture based on graph neural network to finish the joint beam selection and link activation, with taking the network topology information into account. Meanwhile, we present an unsupervised Lagrangian dual learning framework to train the parameters of the GBLinks model. Numerical results show that the proposed GBLinks model can converge to a stable point with the number of iterations increases, in terms of the weighted sum rate. Furthermore, the GBLinks model can reach near-optimal solutions through comparing with the exhaustive scheme in small-scale ultra-dense D2D mmWave communication networks and outperforms GreedyNoSched and the SCA-based method. It also shows that the GBLinks model can generalize to varying network densities and network coverage regions of ultra-dense D2D mmWave communication networks.

In wireless networks, the optimization problems generally have complex constraints and are usually solved via utilizing the traditional optimization methods that have high computational complexity and need to be executed repeatedly with the change of network environments. In this paper, to overcome these shortcomings, an unsupervised deep unrolling framework based on projection gradient descent (PGD), i.e., unrolled PGD network (UPGDNet), is designed to solve a family of constrained optimization problems. The set of constraints is divided into two categories according to the coupling relations among optimization variables and the convexity of constraints. One category of constraints includes convex constraints with decoupling among optimization variables, and the other category of constraints includes non-convex or convex constraints with coupling among optimization variables. Then, the first category of constraints is directly projected onto the feasible region, while the second category of constraints is projected onto the feasible region using a neural network. Finally, an unrolled sum rate maximization network (USRMNet) is designed based on UPGDNet to solve the weighted SR maximization problem for the multiuser ultra-reliable low latency communication system. Numerical results show that USRMNet has a comparable performance with low computational complexity and an acceptable generalization ability in terms of the user distribution.