Drive Test Minimization Using Deep Learning with Bayesian Approximation

Thrane, Jakob; Artuso, Matteo; Zibar, Darko; Christiansen, Henrik Lehrmann

doi:10.1109/vtcfall.2018.8690911

Cited by 19 publications

(19 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [7], Deep Learning (DL) is shown to be capable of inferring radio quality parameters using satellite images and can offer a metric of uncertainty using Bayesian approximation. The aim and novelty of this paper are to investigate the proposed DL method for radio propagation prediction and compare them to existing methodologies.…”

Section: A Contributionsmentioning

confidence: 99%

Model-Aided Deep Learning Method for Path Loss Prediction in Mobile Communication Systems at 2.6 GHz

Zibar²,

2020

Self Cite

View full text Add to dashboard Cite

Accurate channel models are essential to evaluate mobile communication system performance and optimize coverage for existing deployments. The introduction of various transmission frequencies for 5G imposes new challenges for accurate radio performance prediction. This paper compares traditional channel models to a channel model obtained using Deep Learning (DL)-techniques utilizing satellite images aided by a simple path loss model. Experimental measurements are gathered and compose the training and test set. This paper considers path loss modelling techniques offered by state-of-the-art stochastic models and a ray-tracing model for comparison and evaluation. The results show that 1) the satellite images offer an increase in predictive performance by ≈ 0.8 dB, 2) The model-aided technique offers an improvement of ≈ 1 dB, and 3) that the proposed DL model is capable of improving path loss prediction at unseen locations for 811 MHz with ≈ 1 dB and ≈ 4.7 dB for 2630 MHz.

show abstract

Section: A Contributionsmentioning

confidence: 99%

Model-Aided Deep Learning Method for Path Loss Prediction in Mobile Communication Systems at 2.6 GHz

Zibar²,

2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…Novel and simple solutions for determining the LOS-state are of great interest for use with such empirical models. The use of deep learning for learning geographical information from simple data such as satellite images have been demonstrated in [17] and documents improved predictive performance for frequencies at 2.6 GHz.…”

Section: Discussionmentioning

confidence: 99%

Comparison of Empirical and Ray-Tracing Models for Mobile Communication Systems at 2.6 GHz

Thrane

Zibar

Christiansen

2019

2019 IEEE 90th Vehicular Technology Conference (VTC2019-Fall)

Self Cite

View full text Add to dashboard Cite

Accurate channel models for predicting received power under slow fading impairments are essential for planning 5G solutions due to the increased range of possible transmission frequencies. The densification of base stations will pose an increased number of complex coverage and capacity situations where flexible and computational simple channel models are essential. In this paper, we study state-of-the-art empirical channel models, more specifically ITU-R M.2412 and 3GPP 38.901, and their performance on experimental measurements at 2630 MHz for LTE-A reference parameters such as RSRP. A crude raytracing model is implemented for reference. The results show an increase in the predictive performance of approximately 4 dB at 811 MHz compared to higher frequencies of 2630 MHz.

show abstract

“…However, the impact of quantization or sparse MDT measurements is not the focus of these works. Acknowledging the limitations of position estimation methods such as GPS positioning or metrics such as observed time difference of arrival in combination with angles of reception, authors in [19] propose to use big data processing to obtain network performance, such as coverage evaluation as a function of location. To this end, the authors in this work leverage deep neural network and Bayesian probability theory-based techniques to reduce the number of required drive test measurements for LTE networks.…”

Section: ) Positioning Uncertainty Onlymentioning

confidence: 99%

“…However, deep learning-based training requires abundant data and is not likely to work in scenarios with sparse user data. Moreover, the joint impact of quantization and positioning inaccuracy is not the focus of the work in [19]. In contrast, presented work aims to address this challenge of sparse user data along with exploring the trade-off of bin size and positioning accuracy for enhanced MDT-based performance estimation.…”

Section: ) Positioning Uncertainty Onlymentioning

confidence: 99%

“…At each step, a soft-thresholding operation is performed on the singular values of the matrix Q t . By selecting a large value of the parameter, ⌘ in (21), the sequence of iterates, {P t } converges to a matrix which nearly minimizes (19). Starting with Q 0 = 0 2 R (n⇥n) , the algorithm inductively defines…”

Section: ) Rank Minimization Based Matrix Completionmentioning

confidence: 99%

See 1 more Smart Citation

Enhanced MDT-Based Performance Estimation for AI Driven Optimization in Future Cellular Networks

2020

View full text Add to dashboard Cite

Minimization of drive test (MDT) allows coverage estimation at a base station by leveraging measurement reports gathered at the user equipment (UE) without the need for drive tests. Therefore, MDT is a key enabling feature for data and artificial intelligence driven autonomous operation and optimization in future cellular networks. However, to date, the utility of MDT feature remains thwarted by issues such as sparsity of user reports and user positioning inaccuracy. We characterize three key types of errors in MDT-based coverage estimation that stem from inaccurate user positioning, scarcity of user reports and quantization. For the first time, the presented analysis shows existence of joint interplay between these errors on coverage estimation that result from inter-dependency between positioning error and bin width. The analysis also shows that there exists an optimal bin width for given user positioning inaccuracy and user density that minimizes the overall error in MDT-based estimated coverage. Utility of our framework is presented by addressing two applications from network optimization perspective: determining optimal bin width to maximize accuracy of MDT-based coverage estimation and its calibration to further improve its accuracy. INDEX TERMS Minimization of drive test, autonomous coverage estimation, optimal bin width, sparse data, coverage calibration, data driven optimization

show abstract

Drive Test Minimization Using Deep Learning with Bayesian Approximation

Cited by 19 publications

References 9 publications

Model-Aided Deep Learning Method for Path Loss Prediction in Mobile Communication Systems at 2.6 GHz

Model-Aided Deep Learning Method for Path Loss Prediction in Mobile Communication Systems at 2.6 GHz

Comparison of Empirical and Ray-Tracing Models for Mobile Communication Systems at 2.6 GHz

Enhanced MDT-Based Performance Estimation for AI Driven Optimization in Future Cellular Networks

Contact Info

Product

Resources

About