Non-orthogonal multiple access (NOMA) is one of the promising technologies to meet the huge access demand and high data rate requirements of the next generation networks. In this paper, we investigate the joint subchannel assignment and power allocation problem in an uplink multi-user NOMA system to maximize the energy efficiency (EE). Different from conventional model-based resource allocation methods, we propose three deep reinforcement learning (DRL) based frameworks to solve this non-convex optimization problem, referred to as discrete DRL based resource allocation (DDRA) framework, continuous DRL based resource allocation (CDRA) framework and joint DRL and optimization resource allocation (DORA) framework. Specifically, for the DDRA framework, a multi-DQN based network is designed to dynamically allocate resources discretely, which can reduce the output dimension and improve the learning efficiency. To overcome the loss of power discretization in DDRA, a joint DQN and deep deterministic policy gradient (DDPG) based network (CDRA framework) is designed to generate the resource allocation policy. The DORA framework is then proposed as a performance boundary. Finally, an event triggered learning method is combined with all three frameworks to further reduce the computational consumption. Numerical results show that the proposed frameworks can improve the EE performance of the uplink NOMA system and reduce the computation time.

In this paper, we investigate the power allocation for maximizing weighted sum-rate (WSR) in downlink multiple carrier non-orthogonal multiple access (MC-NOMA) systems with imperfect successive interference cancellation (SIC). We formulate the power allocation problem as a non-convex optimization problem with the total power constraint of all sub-channels, while considering often-neglected issues of SIC error and power order constraints at users. Firstly, we discuss the optimization problem assuming receivers can perform perfect SIC, and we provide a concavity condition of the WSR maximization problem for the MC-NOMA system. When the concavity condition is not satisfied, a fractional quadratic transformation is used to overcome the difficulty of problem non-convexity. Based on the transformation, we propose an iterative power allocation algorithm. Then, we consider the SIC error and the power order constraints in the optimization problem, and present a power allocation method with imperfect SIC. Moreover, for both perfect and imperfect SIC, we derive some propositions of the optimal power allocation solution to the WSR maximization problem, and propose a low-complexity power allocation algorithm based on these propositions. Finally, we provide a joint user scheduling and power allocation algorithm for maximizing the WSR. Simulation results illustrate that the proposed resource allocation methods have better performance than the existing schemes.

In this paper, we propose two deep learning (DL) based receiver schemes in uplink multiple-input multiple-output (MIMO) systems. In the first scheme, we design a pilot-assisted MIMO receiver using a data-driven full connected neural network. This data-driven receiver can recover transmitted signal directly in an end-to-end manner without explicitly estimating channel. In the second scheme, we adopt a model-driven network which combines communication knowledge with DL. The model-driven scheme divides the MIMO receiver into channel estimation subnet and signal detection subnet, and each subnet is composed of a traditional solution as initialization and a DL network to further improve the accurate. The simulation results show that both of the two schemes achieve better bit error ratio (BER) performance than traditional methods. In particular, the data-driven scheme can achieve optimal BER performance in low-dimensional MIMO systems, while the model-driven scheme can be trained with fewer trainable parameters and outperforms the data-driven scheme in high-dimension MIMO systems.

Orthogonal time frequency space (OTFS) modulation has been advocated as a promising waveform for achieving integrated sensing and communication (ISAC) due to its superiority in high-mobility adaptability and spectral efficiency. Despite its advantages, due to the complex input-output relation of the OTFS processing, the current OTFS based ISAC schemes require high computation complexity/latency to achieve satisfactory sensing accuracy. In this paper, a concisely fast/slow-time representation for the OTFS input-output relation is derived, which can clearly reveal the physical interactions between the OTFS signals and sensing channels. Under this basis, a novel delay-Doppler spectrum matching assisted active sensing framework is proposed for the OTFS based ISAC systems. The framework consists of three sequentially operated modules, i.e., a coarse searching and triangular spectrum matching module, a local fine tune and quadratic spectrum matching module, and a constant false alarm rate (CFAR) target detection module. The three modules work coordinately to realize joint target detection and high-accuracy target range/velocity estimation. Simulation results show that the proposed framework achieves better detection performance and range/velocity estimation accuracy with much less computations when compared with the reference schemes.

In this paper, we study the edge caching problem by considering the heterogeneous context with unknown users’ preferences. The Cache Provider (CP) can personalize the users’ storage based on available data to maximize the overall cache hit rate, accounting for the dynamic natures of both mobile edge cache scenarios and users’ preferences. Toward this end, we introduce an online Bayesian clustering caching algorithm for the CP to autonomously learn the users’ interactive cache hit data in a collaborative way while maintaining sustainable scalability. Specifically, a Bayesian generative framework called the Dirichlet Multinomial Mixture (DMM) model is used to describe the uncertainty about the latent number of users’ clusters, each of which consists of the users with the same preference. Then, a dynamic clustering policy is proposed to obtain both the underlying mapping of users to clusters and the preferences of each cluster by using a collapsed Gibbs Sampling algorithm. Subsequently, cache decisions are made according to the generated mappings by extending the traditional cache bandit algorithm to a new bandit mechanism with clusters of arms, capable of expediting the learning process between exploitation and exploration. We theoretically characterize the value of dynamic Bayesian clustering for the long-term edge caching scenario with respect to the regret incurred by non-cluster schemes. Finally, using a real-world dataset, our numerical results show that the proposed scheme outperforms the caching algorithms without clustering in the uncertain network scenario.

Intelligent reflecting surface (IRS)-enhanced wireless communication is one of the critical technologies of the sixth generation mobile communication, which has recently emerged as an innovative solution to the realization of intelligent transportation. It can meet the increasing demand for high-speed railway (HSR) communications services. In this paper, we consider a train-ground TDD wireless mobile communication paradigm to deploy two IRSs on the roadside and the train roof to assist the wireless transmission for the HSR communication system. To achieve the more significant performance gain, in the uplink transmission phase, we propose the practical channel estimation framework of the MIMO system aided by double IRSs to acquire reliable channel state information (CSI) of each channel. To facilitate the processing of the three-dimension tensor, we leverage the trilinearity of the parallel factor decomposition to decompose the received signal into a two-dimensional matrix. The Doppler frequency offset compensating scheme and channel tracking scheme based on Kalman-Filter are proposed to further decrease the pilot overheads. Then, in the downlink phase, a block coordinate descent (BCD) algorithm is proposed to handle the data rate maximization problem of double IRSs communication system with the estimated CSI. We reformulate the original problem and decouple the double IRSs phase optimization problem into two single IRS phase optimization non-convex problems under the unit-modulus constraint. The Majorization-Minimization (MM) algorithm is proposed to approximate the optimal solution of the two optimization subproblems. Simulation results corroborate the effectiveness of the proposed channel estimation and phase shifts optimization scheme.