1,721,039 research outputs found
Deep Learning for Direction of Arrival Estimation in Massive MIMO Systems
No full textWith the dawn of fifth generation (5G) cellular networks massive MIMO systems have become more applicable. One problem in massive MIMO systems is pilot contamination caused by the dense placement of 5G cellular towers. Therefore it is of interest to use blind source separation instead to determine channel parameters. Traditional methods for blind source separation can be slow so alternatives are of interest. The three methods examined are deep neural networks, sparse estimation methods (ISTA based) and a fusion between the two (Learned-ISTA based). For direction of arrival estimation the methods provide comparable NMSE at the low end of 10e-1 to high end of 10e-2. The Learned-ISTA based methods show promise and further investigation of training it, model parameters and peak-finding algorithms for selecting the directions of arrival is relevant
Federated Interference Management for Industrial 6G Subnetworks
No full text6G in-X subnetværk er kortdistances laveffektsceller, der er designet til at imødekomme ekstreme krav til kommunikation i form af datahastighed, forsinkelse og pålidelighed. Dog udgør interferens en betydelig begrænsende faktor for ekstrem kommunikation i tætte udrulninger af in-X subnetværk. Nylige studier har foreslået løsninger til styring af interferens baseret på ”multi-agent reinforcement learning”, hvor problemet med radioressourceoptimering modelleres som en ”multi-Markov decision process”. Studierne har været baseret på enten centraliseret eller distribueret træning. Mens centraliseret træning drager fordel af erfaringerne fra alle delnetværkene under træningen, kan det medføre kompromittering af privatliv og sikkerhed, da det kræver deling af målinger mellem delnetværkene og en centraliseret agent. Derimod er agenter i distribueret træning udelukkende afhængige af lokale målinger af miljøet for at træffe beslutninger, hvilket ofte fører til konvergensproblemer. For at overvinde disse udfordringer foreslås en klient-til-server ”horizontal federated reinforcement learning” model, hvor viden deles implicit gennem lokalt trænede modelvægte. Simulationer i et industrielt miljø ved hjælp af 3GPP-udbredelsesmodeller har vist lovende resultater med hensyn til hurtig konvergens, marginale præstationsforbedringer og robusthed over for ikke-stationære miljøer.6G in-X subnetworks are short-range low-power cells envisioned to support extreme communication requirements for data rate, latency, and reliability. However, interference represents a major limiting factor to extreme communication in dense deployments of in-X subnetworks. Recent studies have proposed interference management solutions based on multi-agent reinforcement learning, where the radio resource optimization problem is modeled as a multi-Markov decision process. The studies have been based on centralized or distributed training. While centralized training benefits from the experiences of all subnetworks during the training, it may lead to compromised privacy and security issues since it requires sharing of measurements between the subnetworks and a centralized agent. In contrast, agents in distributed training rely solely on only local measurements of the environment for decision which often leads to convergence problems.To overcome these challenges, a client-to-server horizontal federated reinforcement learning framework is proposed, where knowledge is shared implicitly through locally trained model weights.Simulations in an industrial environment using 3GPP propagation models have shown promising results for quick convergence, marginal performance improvement, and robustness to non-stationary environments
Machine Learning and Artificial Intelligence Enabled Failure Prediction for Maritime Propulsion Systems
No full textThis thesis investigates the application of machine learning (ML) and artificial intelligence(AI) techniques for failure prediction in maritime propulsion systems, with a focus on twocritical failures observed in MAN Energy Solutions’ engines: accumulator membrane fail-ure in sequential two-stroke engines and pipe damage during methanol fuel changeover indual-fuel systems. The thesis explores the feasibility of predictive maintenance throughdata-driven approaches, leveraging time series data collected from operational vessels andcontrolled test engines. A comprehensive feature engineering pipeline is developed to ad-dress challenges related to high dimensionality, feature redundancy, and class imbalance.Feature selection methods are applied to construct interpretable and task-relevant fea-ture subsets, resulting in insights into system behavior leading up to failure. Multiple MLand AI models, including Random Forests, 1D Convolutional Neural Networks (CNNs),and Long Short-Term Memory (LSTM) networks, are implemented and evaluated forboth binary failure classification and remaining useful life (RUL) regression tasks, fordifferent reading window (RW) and prediction window (PW) sizes. Here SMOTE andDTW-SMOTE has been used to augment data, increasing the data variance in training.Results demonstrate that methanol changeover failures are highly predictable, with Ran-dom Forest and 1D CNN models achieving strong performance with an accuracy of 91%for changeover classification and an R2 of 0.826 for changeover RUL prediction. In con-trast, accumulator failure prediction proves more challenging due to limited failure dataand lower signal resolution with the 1D CNN providing the greatest accuracy of 71%.The findings highlight the importance of model selection, task-specific window sizing, andfeature interpretability in predictive maintenance applications. The thesis concludes thatML and AI can be used for predicting pipe damage during methanol fuel changeover,while providing valuable insight into the system and the failure through feature engineer-ing. This task is slightly harder for accumulator membrane failure in sequential two-strokeengines, and further work is required to develop high performing and robust models forthis purpose.<br/
Channel Prediction for Mobile MIMO Wireless Communication Systems
No full textTemporal variation and frequency selectivity of wireless channels constitute
a major drawback to the attainment of high gains in capacity
and reliability offered by multiple antennas at the transmitter and receiver
of a mobile communication system. Limited feedback and adaptive transmission
schemes such as adaptive modulation and coding, antenna selection,
power allocation and scheduling have the potential to provide the platform
of attaining the high transmission rate, capacity and QoS requirements in
current and future wireless communication systems. Theses schemes require
both the transmitter and receiver to have accurate knowledge of Channel
State Information (CSI). In Time Division Duplex (TDD) systems, CSI at
the transmitter can be obtained using channel reciprocity. In Frequency Division
Duplex (FDD) systems, however, CSI is typically estimated at the
receiver and fed back to the transmitter via a low-rate feedback link. Due to
the inherent time delays in estimation, processing and feedback, the CSI obtained
from the receiver may become outdated before its actual usage at the
transmitter. This results in significant performance loss, especially in high
mobility environments. There is therefore a need to extrapolate the varying
channel into the future, far enough to account for the delay and mitigate the
performance degradation.
The research in this thesis investigates parametric modeling and prediction
of mobile MIMO channels for both narrowband and wideband systems.
The focus is on schemes that utilize the additional spatial information offered
by multiple sampling of the wave-field in multi-antenna systems to
aid channel prediction. The research has led to the development of several
algorithms which can be used for long range extrapolation of time-varyingchannels. Based on spatial channel modeling approaches, simple and efficient
methods for the extrapolation of narrowband MIMO channels are proposed.
Various extensions were also developed. These include methods for wideband
channels, transmission using polarized antenna arrays, and mobile-to-mobile
systems.
Performance bounds on the estimation and prediction error are vital when
evaluating channel estimation and prediction schemes. For this purpose, analytical
expressions for bound on the estimation and prediction of polarized
and non-polarized MIMO channels are derived. Using the vector formulation
of the Cramer Rao bound for function of parameters, readily interpretable
closed-form expressions for the prediction error bounds were found for cases
with Uniform Linear Array (ULA) and Uniform Planar Array (UPA). The
derived performance bounds are very simple and so provide insight into system
design.
The performance of the proposed algorithms was evaluated using standardized
channel models. The effects of the temporal variation of multipath
parameters on prediction is studied and methods for jointly tracking the
channel parameters are developed. The algorithms presented can be utilized
to enhance the performance of limited feedback and adaptive MIMO
transmission schemes
Capacity and Error Rate Analysis of MIMO Satellite Communication Systems in Fading Scenarios
Get PDFIn this paper, we investigated the capacity and bit error rate (BER) performance of Multiple Input Multiple Output (MIMO) satellite systems with single and multiple dual polarized satellites in geostationary orbit and a mobile ground receiving station with multiple antennas. We evaluated the effects of both system parameters such as number of satellites, number of receive antennas, and SNR and environmental factors including atmospheric signal attenuations and signal phase disturbances on the overall system performance using both analytical and spatial models for MIMO satellite systems.DOI:http://dx.doi.org/10.11591/ijece.v4i4.534
Mathematical Modelling and Prediction of Interference Power in In-robot Subnetworks
No full textThe envisioned wireless networks, 6G in-X subnetworks have extreme requirements for latency, data rate and reliability. The in-X subnetworks considered in this thesis, is the in-robot subnetworks in a factory setting. The interference power which comes from the in-robot subnetworks in a factory setting is analyzed and predicted throughout this thesis.Analytical expressions for the mean and auto-correlation function (ACF) of the interference power are derived. The expressions for the mean and ACF contained high dimensional integrals which we estimate using Monte Carlo (MC) integration. When estimating the expression for the mean it was found that the analytical mean fitted the mean of the interference power obtained from simulations. It was found that when estimating the expression for the ACF using MC integration that it would overestimate the ACF of simulated interference power. However, the ACFs had the same form, thus, an appropriate scaling could make them coincide.Furthermore, the interference power simulated from in-robot subnetworks was predicted using an autoregressive (AR) model of order 20. We also found that the interference power could be predicted using the AR(20) predictor for up to 8 [ms] when the velocity is 2 [m/s]. Additionally, the AR(20) predictor outperformed the last value predictor for all settings
Unsupervised Deep Unfolded PGD for Transmit Power Allocation in Wireless Systems
Get PDFTransmit power control (TPC) is a key mechanism for managing interference, energy utilization, and connectivity in wireless systems. In this paper, we propose a simple low-complexity TPC algorithm based on the deep unfolding of the iterative projected gradient descent (PGD) algorithm into layers of a deep neural network and learning the step-size parameters. An unsupervised learning method with either online learning or offline pretraining is applied for optimizing the weights of the DNN. Performance evaluation in dense device-to-device (D2D) communication scenarios showed that the proposed method can achieve better performance than the iterative algorithm with more than a factor of 2 lower number of iterations.</p
Calibration of Stochastic Radio Propagation Models Using Machine Learning
Get PDFThis letter proposes a machine learning based method for the calibration of stochastic radio propagation models. Model calibration is cast as a regression problem involving mapping of the channel transfer function or impulse response to the model parameters. A multilayer perceptron is trained with summary statistics computed from synthetically generated channel realizations using the model. To calibrate the model, the trained network is used to estimate the model parameters from channel statistics obtained from measurements. The performance of the proposed method is evaluated with propagation graph and Saleh-Valenzuela models using both simulated data and in-room channel measurements. Results show accurate estimation of the parameters of both models
Federated Multi-Agent DRL for Radio Resource Management in Industrial 6G in-X subnetworks
Get PDFRecently, 6G in-X subnetworks have been proposed as low-power short-range
radio cells to support localized extreme wireless connectivity inside entities
such as industrial robots, vehicles, and the human body. Deployment of in-X
subnetworks within these entities may result in rapid changes in interference
levels and thus, varying link quality. This paper investigates distributed
dynamic channel allocation to mitigate inter-subnetwork interference in dense
in-factory deployments of 6G in-X subnetworks. This paper introduces two new
techniques, Federated Multi-Agent Double Deep Q-Network (F-MADDQN) and
Federated Multi-Agent Deep Proximal Policy Optimization (F-MADPPO), for channel
allocation in 6G in-X subnetworks. These techniques are based on a
client-to-server horizontal federated reinforcement learning framework. The
methods require sharing only local model weights with a centralized gNB for
federated aggregation thereby preserving local data privacy and security.
Simulations were conducted using a practical indoor factory environment
proposed by 5G-ACIA and 3GPP models for in-factory environments. The results
showed that the proposed methods achieved slightly better performance than
baseline schemes with significantly reduced signaling overhead compared to the
baseline solutions. The schemes also showed better robustness and
generalization ability to changes in deployment densities and propagation
parameters.Comment: Accepted for Workshop on Industrial Wireless Networks - IEEE
International Symposium on Personal, Indoor and Mobile Radio Communication
- …
