Bin Li

As various modulation techniques have being proposed to meet the practical demands of high-mobility communications, this paper introduces a unified modulation framework based on fast unitary transformations, drawing inspiration from the use of diversity to combat selective fading. We show that several classic modulation schemes arise as special cases of the proposed framework and devise a unified implementation architecture. The unified framework enables us to leverage the unique characteristics of each domain to optimize system performance and streamline the signal processing workflow. Finally, we explore how the proposed unified modulation can be integrated with some key technologies in 6G, paving the way for the evolution of future communication networks.

Tensor decomposition (TD) models are promising solutions for knowledge graph completion due to their simple structures but powerful representation capacities. The TD models typically adopt Tucker decomposition with a structured core tensor. Some models with a sparse core tensor, such as DistMult and ComplEx, are too simple and thus limit the interaction between embedding components, while other models with a dense core tensor are too complex and may lead to significant overfitting. To address these issues, we propose a new TD model called SPAC (Sparse Partitioning and Adaptive Core tensor pruning) model for knowledge graph completion. Specifically, SPAC captures coarse and fine-grained semantic information using a hybrid core tensor, where auxiliary cores are used to model sparse interactions and main cores for dense interactions. Moreover, SPAC introduces a gating mechanism to control the output of intermediate variables, enhancing the interaction between different partition groups. Furthermore, SPAC employs an adaptive pruning approach to dynamically adjust the shape of the core tensor. Due to the elaborate model design, the proposed TD model enhances expressive capacity and reduces the number of parameters in the core tensor. Experiments are conducted on datasets FB15k-237, WN18RR, and YAGO3-10. The results demonstrate that SPAC outperforms state-of-the-art tensor decomposition models, including MEIM and Tucker models. A series of ablation studies show that the gating mechanism and adaptive pruning strategy in SPAC are crucial for the performance improvement.

The emerging field of graph learning, which aims to learn reasonable graph structures from data, plays a vital role in Graph Signal Processing (GSP) and finds applications in various data processing domains. However, the existing approaches have primarily focused on learning deterministic graphs, and thus are not suitable for applications involving topological stochasticity, such as epidemiological models. In this paper, we develop a hierarchical Bayesian model for graph learning problem. Specifically, the generative model of smooth signals is formulated by transforming the graph topology into self-expressiveness coefficients and incorporating individual noise for each vertex. Tailored probability distributions are imposed on each edge to characterize the valid graph topology constraints along with edge-level probabilistic information. Building upon this, we derive the Bayesian Graph Learning (BGL) approach to efficiently estimate the graph structure in a distributed manner. In particular, based on the specific probabilistic dependencies, we derive a series of message passing rules by a mixture of Generalized Approximate Message Passing (GAMP) message and Belief Propagation (BP) message to iteratively approximate the posterior probabilities. Numerical experiments with both artificial and real data demonstrate that BGL learns more accurate graph structures and enhances machine learning tasks compared to state-of-the-art methods.

By deploying the ubiquitous and reliable coverage of low Earth orbit (LEO) satellite networks using optical inter satellite link (OISL), computation offloading services can be provided for any users without proximal servers, while the resource limitation of both computation and storage on satellites is the important factor affecting the maximum task completion time. In this paper, we study a delay-optimal multi-satellite collaborative computation offloading scheme that allows satellites to actively migrate tasks among themselves by employing the high-speed OISLs, such that tasks with long queuing delay will be served as quickly as possible by utilizing idle computation resources in the neighborhood. To satisfy the delay requirement of delay-sensitive task, we first propose a deadline-aware task scheduling scheme in which a priority model is constructed to sort the order of tasks being served based on its deadline, and then a delay-optimal collaborative offloading scheme is derived such that the tasks which cannot be completed locally can be migrated to other idle satellites. Simulation results demonstrate the effectiveness of our multi-satellite collaborative computation offloading strategy in reducing task complement time and improving resource utilization of the LEO satellite network.

Fingerprint-based indoor localization holds great potential for the Internet of Things. Despite numerous studies focusing on its algorithmic and practical aspects, a notable gap exists in theoretical performance analysis in this domain. This article aims to bridge this gap by deriving several lower bounds and approximations of mean square error (MSE) for fingerprint-based localization. These analyses offer different complexity and accuracy tradeoffs. We derive the equivalent Fisher information matrix and its decomposed form based on a wireless propagation model, thus obtaining the Cramér–Rao bound (CRB). By approximating the Fisher information provided by constraint knowledge, we develop a constraint-aware CRB. To more accurately characterize nonlinear transformation and constraint information, we introduce the Ziv-Zakai bound (ZZB) and modify it for adapt deterministic parameters. The Gauss–Legendre quadrature method and the trust-region reflective algorithm are employed to make the calculation of ZZB tractable. We introduce a tighter extrapolated ZZB by fitting the quadrature function outside the well-defined domain based on the $Q$ -function. For the constrained maximum likelihood estimator, an approximate MSE (AMSE) expression, which can characterize map constraints, is also developed. The simulation and experimental results validate the effectiveness of the proposed bounds and AMSE.

Message passing algorithms on graphical models offer a low-complexity and distributed paradigm for performing marginalization from a high-dimensional distribution. However, the convergence behaviors of message passing algorithms can be heavily affected by the adopted message update schedule. In this paper, we focus on the variational message passing (VMP) applied to Gaussian graphical models and its convergence under different schedules is analyzed. In particular, based on the update equations of VMP under the mean-field assumption, we prove that the mean vectors obtained from VMP are the exact marginal mean vectors under any valid message passing schedule, giving the legitimacy of using VMP in Gaussian graphical models. Furthermore, three categories of valid message passing schedules, namely serial schedule, parallel schedule and randomized schedule are considered for VMP update. In the basic serial schedule, VMP unconditionally converges, but could be slow in large-scale distributed networks. To speed up the serial schedule, a group serial schedule is proposed while guaranteeing the VMP convergence. On the other hand, parallel schedule and its damped variant are applied to accelerate VMP, where the necessary and sufficient convergence conditions are derived. To allow nodes with different local computation resources to compute messages more flexibly and efficiently, a randomized schedule is proposed for VMP update, and the probabilistic necessary and sufficient convergence conditions are presented. Finally, numerical results and applications are presented to illustrate the trade-offs in the ease and speed of convergence.

Indoor localization is of pivotal significance for a wide variety of services in the context of the Internet of Things (IoT). Both ranging-based and fingerprint-based localization techniques are promising for employment in harsh indoor environments. Hence, we propose a unified framework based on factor graphs for ubiquitous high-accuracy indoor localization. Our unified framework efficiently integrates ranging and fingerprinting for striking an appealing accuracy versus deployment cost tradeoff, where the crowdsourcing required for the construction of fingerprinting databases can also be addressed with little human intervention. By intrinsically amalgamating the global grid sampling and the regularized importance-resampling techniques, a nonparametric belief propagation algorithm is proposed for achieving the accurate position estimation at the cost of a moderate computational complexity. For improving the robustness to environmental variations, a likelihood-ratio-based approach is employed to detect ranging outliers. Moreover, a low-complexity serial scheduling scheme defined over factor graphs is designed for real-time localization. We design a hybrid ultrawide bandwidth and Wi-Fi localization system relying on off-the-shelf commercial devices and evaluate the proposed unified framework in a typical office building. Our experimental results show that the proposed algorithm outperforms the existing state-of-the-art methods and it is capable of achieving submeter localization accuracy.

Satellite communication is capable of supporting seamless global coverage. However, owing to the reliance on limited-duration solar power, the high power amplifier (HPA) is often driven close to its saturation point, which leads to severe nonlinear distortion in satellite channels. Thus, mitigating the effect of the nonlinear distortion becomes essential for reliable communications. In this paper, we propose an efficient joint channel estimation and data detection method based on message passing within the associated factor graph modelling the HPA employed in nonlinear satellite channels. Then, we develop a combined belief propagation and mean field (BP-MF) method to cope with the hard constraints and dense short loops on the factor graph. In particular, the parametric message updating expressions relying on the canonical parameters are derived in the symbol detection part. To alleviate the impact of dense loops, we reformulate the system model into a compact form within the channel estimation part and then reconstruct a loop-free subgraph associated with vector-valued nodes to guarantee convergence. Furthermore, the proposed BP-MF method is also extended to the realistic scenario of having unknown noise variance. To further reduce the computational complexity of the large-scale matrix inversion of channel estimation, the generalized approximate message passing (GAMP) algorithm is employed to decouple the vector of channel coefficient estimation into a series of scalar estimations. Simulation results show that the proposed methods outperform the state-of-the-art benchmarks both in terms of bit error rate performance and channel estimation accuracy.

Belief propagation (BP) is a popular message passing algorithm for distributed cooperative localization. However, due to the nonlinearity of measurement functions, BP implementation has no closed-form expression and requires message approximations. While nonparametric BP can be used, it suffers from a high computational complexity, thus being impractical in energy-constrained networks. In this paper, a parametric Bayesian method with Gaussian BP implementation is proposed for distributed cooperative localization. With linearization of the Euclidean norm in ranging measurements, the joint posterior distribution of agents’ locations is successively approximated with a sequence of high-dimensional Gaussian distributions. At each iteration of the successive Gaussian approximation, vector-valued Gaussian BP is further adopted to compute the marginal distributions of agents’ locations in a distributed way. It is proved by the principle of majorization-minimization that the proposed successive Gaussian approximation is guaranteed to converge, and the sequence of the estimated agents’ locations converges to a stationary point of the objective function of the maximum a posteriori estimation. Furthermore, although cooperative localization involves loopy network topologies, in which convergence property of Gaussian BP is generally unknown, it is proved in this paper that vector-valued Gaussian BP converges, making the proposed parametric BP-based method being the first one achieving convergence guarantee. Compared to the nonparametric BP counterpart, the proposed method has a much lower computational complexity and communication overhead. Simulation results demonstrate that the proposed method achieves a superior performance in localization accuracy compared to existing cooperative localization methods.