Wireless channel scenario identification carries importance as it assists in network design aspects such as receiver modeling and channel prediction. In this work, we consider the identification of indoor and outdoor scenarios using convolutional neural networks (CNNs) with channel measurements as the input data. First, we consider the channel measurements obtained as independent realizations (the snapshot dataset) and utilize a 2-D CNN to classify between indoor and outdoor channels. Then, the channel measurements taken for successive time samples (the time series dataset) are classified using 3-D CNNs. Finally, comparison of the CNN-based approaches is performed with a traditional machine learning-based k-nearest neighbors (k-NN) algorithm, which takes skewness, kurtosis, and delay spread of the channel data as the input features. The CNN-based method for the snapshot data yields around 87% accuracy and performs better than the baseline k-NN method yielding around 76%. For the time series dataset, the CNN-based method with up to a 85% accuracy shows a marked improvement over the 60% accuracy achieved by the k-NN algorithm.

Outdated and inaccurate channel state information (CSI) prevents the transmitter from adapting to current channel conditions and impairs reliable communication. This paper proposes a deep learning-based methodology for real-time prediction of future CSI, which enables it to be fed back in advance and mitigates the effect of the feedback and processing delay. Predicting the future CSI at the user equipment may also help the gNodeB (gNB) send reference signals less frequently and may help reduce the reference signal transmission overhead. Our approach utilizes Long Short-Term Memory (LSTM) type recurrent neural networks (RNN) to predict the CSI from past channel estimations. However, the knowledge of past channel gains is assumed to be limited, and only partial channel data is available. We propose a new algorithm that utilizes LSTMs and remedies incomplete channel history problem, which we denote by RNN-I. We evaluate the performance of the proposed framework by comparing it with two conventional techniques: persistence and auto-regressive (AR) model-based prediction methods. The results show that the proposed RNN-I algorithm performs well under strong channel mismatch scenarios, while AR methods diverge and can not keep up with fast varying channels if the mismatch in the channel characteristics used in training and inference is significant.

In this paper, we consider the achievable sum-rate/distortion tradeoff for the Gaussian central estimation officer (CEO) problem with a scalar source having arbitrary memory. We describe how the arbitrary memory problem can be fully characterized by using known results for the vector CEO problem, and then we formulate the variational problem of minimizing the sum-rate subject to a distortion constraint. To solve the problem, we extend the conventional Lagrange method and show that if the solution exists, it should consist of a zero part and a non-zero part, where the non-zero part is determined by solving a set of Euler equations. By calculating the second variation of the min-sum-rate problem, a sufficient condition is also found that can be used to determine if the necessary solution results in the minimal sum rate. The special case of two terminals is examined in detail, and it is shown that an analytical solution is possible in this case. Analysis and discussion with examples are provided to illustrate the theoretical results. The general solution obtained in this paper is shown to be compatible with the previous results for cases such as the problem of rate evaluation for sources without memory.

In recent years, radio positioning has received increasing attention and found many applications in various areas. However, the existence of non-line-of-sight (NLOS) paths introduces considerable positioning errors. In this paper, we propose a two-step approach in order to deal with pure NLOS scenarios based on a simple assumption regarding the propagation environment. A nonlinear least squares (NLS) method is proposed for the initial estimation, followed by a Kalman filter-based method to track subsequent movements. Compared with previous studies, fewer measurements are required to be made. Simulation results are provided to show the performance of both methods.

In this paper, we consider a dynamic linear system in statespace form where the observation equation depends linearly on a set of parameters. We address the problem of how to dynamically calculate these parameters in order to minimize the mean-squared error (MSE) of the state estimate achieved by a Kalman filter. We formulate and solve two kinds of problems under a quadratic constraint on the observation parameters: minimizing the sum MSE (Min-Sum-MSE) or minimizing the maximum MSE (Min-Max-MSE). In each case, the optimization problem is divided into two sub-problems for which optimal solutions can be found: a semidefinite programming (SDP) problem followed by a constrained least-squares minimization. A more direct solution is shown to exist for the special case of a scalar observation; in particular, the Min-Sum-MSE problem is optimally solved utilizing Rayleigh quotient, and the Min-Max-MSE problemreduces to an SDP feasibility test that can be solved via the bisection method.

In wireless sensor networks, many monitoring problems can be cast in the form of distributed estimation. If the data links from the sensor nodes to the fusion center have limited capacity, there is a tradeoff between estimation precision and transmission rate. This kind of decentralized estimation system is equivalent to the so-called indirect multiterminal source coding problem, and the Berger-Tung inner bound is the best known achievable rate region boundary. In this paper, we attempt to evaluate the Berger-Tung sum rate for a vector source with correlated components. First we formulate the sum rate as a nonconvex optimization problem with a distortion constraint. Then we propose a method to find an approximate solution. Numerical experiments show the approximate solution is accurate if the required distortion level is relatively small. Its appropriateness is also verified by simulation results based on practical quantizer design.

We consider a distributed sensor network in which the single antenna sensor nodes observe a deterministic unknown parameter and after encoding the observed signal with a phase parameter, the sensor nodes transmit it simultaneously to a multi-antenna fusion center (FC). The FC optimizes the phase encoding parameter and feeds it back to the sensor nodes such that the variance of estimation error can be minimized. We relax the phase optimization problem to a semidefinite programming (SDP) problem and the numerical results show that the performance of the proposed method is close to the theoretical bound. Also, asymptotic results show that when the number of sensors is very large and the variance of the distance between the sensor nodes and FC is small, multiple antennas do not provide a benefit compared with a single antenna system; when the number of antennas M is large and the measurement noise at the sensor nodes is small compared with the additive noise at the FC, the estimation error variance can be reduced by a factor of M.