Resource-efficient Transmission of Vehicular SensorData Using Context-aware Communication

Benjamin Sliwa1, Thomas Liebig2, Robert Falkenberg1, Johannes Pillmann1 and Christian Wietfeld1

1Communication Networks Institute, 2Department of Computer Science VIIITU Dortmund University, 44227 Dortmund, Germany

e-mail: {Benjamin.Sliwa, Thomas.Liebig, Robert.Falkenberg, Johannes.Pillmann, Christian.Wietfeld}@tu-dortmund.de

Abstract—Upcoming Intelligent Traffic Control Systems(ITSCs) will base their optimization processes on crowdsensingdata obtained for cars that are used as mobile sensor nodes.In conclusion, public cellular networks will be confronted withmassive increases in Machine-Type Communication (MTC) andwill require efficient communication schemes to minimize theinterference of Internet of Things (IoT) data traffic with humancommunication. In this demonstration, we present an OpenSource framework for context-aware transmission of vehicularsensor data that exploits knowledge about the characteristics ofthe transmission channel for leveraging connectivity hotspots,where data transmissions can be performed with a high gradeif resource efficiency. At the conference, we will present themeasurement application for acquisition and live-visualizationof the required network quality indicators and show how thetransmission scheme performs in real-world vehicular scenariosbased on measurement data obtained from field experiments.

I. INTRODUCTION

With massive deployment of cars used as mobile sensornodes proving information for crowdsensing-based cloud ser-vices, the efficiency of the communication becomes a majorchallenge as this kind of MTC is usually performed in publiccellular networks, leading to a competition of IoT devices andhuman participants among the available network resources.Usually, the obtained data is transmitted in a periodical waywithout considering the properties of the radio channel. Conse-quently, many transmissions happen during low quality periodsof the channel and waste spectrum- and energy resources. Toovercome this issue, the anticipatory communication schemehas been proposed [1]. Instead of immediately transmittingthe measured information, data is aggregated locally if thecurrent channel quality is low and transmitted when the vehicleexperiences a connectivity hotspots, where transmissions canbe performed in a very resource-efficient way. While thisapproach introduces another delay into the processing chaindue to the local buffering, it satisfies the requirements of mostcrowdsensing-based services (e.g. traffic control, road qualityestimation and weather forecast), as these work well with datathat is up-to-date in the dimension of several minutes.

In recent work [2], we proposed a context-aware data trans-mission scheme as an extension to Channel-aware Transmis-sion (CAT) [3] that uses machine learning based throughputprediction in order to optimize the transmission time andavoid transmissions during low channel quality periods. In thispaper, the software framework for measuring and evaluation

of the context-aware transmission schemes is demonstrated.The framework itself is available as Open Source.

II. SYSTEM ARCHITECTURE

Fig. 1 shows the overall system architecture model of theproposed framework used to realize context-aware anticipatorycommunication mechanisms. Instead of sending data in aperiodical way, a probability for the transmission of the wholedata buffer is computed within each time step based on theknowledge about the context characteristics. The schemesoperates at the application level and can therefore easilyintegrating into existing sensor applications. The channelcontext parameters are defined by the Long Term Evolution(LTE) downlink indicators Reference Signal Received Power(RSRP), Reference Signal Received Quality (RSRQ), Signal-to-noise-ratio (SNR) and Channel Quality Indicator (CQI).Furthermore, the mobility context parameters consisting of thevehicle’s velocity and direction as well as position informationand map knowledge. All context parameters are recordedduring the transmission phases and form the training data forthe machine learning based prediction mechanism. In order totake the application requirements into account, the behavior

RS

RP

RS

RQ

SN

RC

QICh

an

ne

lC

on

tex

t Passive ProbingLTE Downlink

Da

tara

teP

aylo

ad

Active ProbingHTTP POST

Po

sitio

nV

elo

city

Dire

ctio

n

Mo

bil

ity

Co

nte

xt GPS Module

Sy

ste

mP

ara

me

ters

Of inefl Training

Online Datarate Prediction

Trained Model

Context-aware Transmission Scheme

Transmission Probability

Metric Current Datarate

Fig. 1. Architecture model for the proposed context-aware communicationframework. The abstract metric definition allows the application of thetransmission scheme to different communication technologies and can be usedto combine multiple transmission metrics.

Accepted for presentation in: 19th IEEE International Conference on Mobile Data Management (MDM)

c© 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses,including reprinting/republishing this material for advertising or promotional purposes, collecting new collected worksfor resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.

is configured with the system parameters that describe thetransmission metric Φ, which is either directly mapped to oneof the downlink indicators or is derived by a combinationof multiple indicators. In [2], we evaluated the data rateprediction accuracy with different machine learning modelsand used the prediction results themselves as a CAT metric.Fig. 2 shows key results for the achieved prediction accuracyas well as for the the machine learning based CAT in areal-world evaluation. Using the data rate prediction as ametric for context-aware communication, the resulting meandata rate has been improved by up to 164% on highwayscenarios. The results further showed that no single networkquality indicator provides enough information on its ownfor identifying a connectivity hotspot with a high grade ofcertainty. Consequently, the machine learning based metric(M5T decision tree) achieves the highest data rate gains asit considers the available information as a whole and sets intorelation to the payload size.

III. DEMONSTRATION

The demonstration at the conference venue will show themain features the context-aware transmission approach andwill feature live capturing of network quality data as well ascontext-aware vehicular communication based on experimentaltraces.

A. Data Acquisition and Live Analysis with an Android-basedMeasuring App

The first part of the demo will address the topic of ob-taining network quality data in real world cellular networks.For this purpose, an Android-based measuring applicationwill be presented that is able to capture and visualize thebehavior of the current network quality indicators as well asperforming and evaluating data transmissions with the CATcommunication scheme. The application is freely availableas Open Source1 and is intended to be used to capture therequired training data for the machine-learning based data rateprediction. Additionally, it serves as a starting point for theevaluation of novel context-aware transmission schemes.

1Available at https://github.com/BenSliwa/MTCApp

Predicted Goodput [MBit/s]

0

2

4

6

8

10

12

14

16

18

20

Measure

d G

oodput [M

Bit/s

]

0 2 4 6 8 10 12 14 16 18 20

(a)Transmission Scheme

0

5

10

15

20

25

Go

od

pu

t [M

Bit/s

]

Periodic SINR CATPes CATM5T,15 CATM5T,18

Controllable trade-off betweengoodput and sensor data age

CAT

(b)

Fig. 2. Accuracy of the machine learning based data rate prediction andresource-efficiency improvements by leveraging the prediction as a metric forcontext-aware communication Source: Adapted from [2]

0 1000 2000 3000 4000 5000 6000 7000 8000

0

20

40

SN

R [dB

]

0 1000 2000 3000 4000 5000 6000 7000 8000-120

-100

-80

RS

RP

[dB

m]

0 1000 2000 3000 4000 5000 6000 7000 8000

-15

-10

-5

RS

RQ

[dB

]

0 1000 2000 3000 4000 5000 6000 7000 8000

Driven distance [m]

0

10

CQ

I

Fig. 3. Live-visualization of the mobility behavior and network qualityindicators based on real-world trace data

B. Visualization of the Real-world Context-aware Transmis-sion Behavior in a Vehicular Scenario

The second part will show how the proposed resource-efficient transmission scheme behaves in a real-world vehicularcontext. For this purpose, we will utilize the obtained tracesfrom the large-scale field evaluation and visualize the dynamicmobility behavior of the measurement vehicle on its driventrajectory as well as the channel conditions the vehicle isexperiencing at the respective locations. The dataset consistsof Key Performance Indicator (KPI) measurements from 4000transmissions that were performed in a public cellular networkwith more than 1000 km total driven distance on is publiclyavailable at [4].

ACKNOWLEDGMENT

Part of the work on this paper has been supported byDeutsche Forschungsgemeinschaft (DFG) within the Collabo-rative Research Center SFB 876 “Providing Information byResource-Constrained Analysis”, projects A4 and B4 andhas been conducted within the AutoMat (Automotive BigData Marketplace for Innovative Cross-sectorial Vehicle DataServices) project, which received funding from the Euro-pean Union’s Horizon 2020 (H2020) research and innova-tion programme under the Grant Agreement No 644657.Thomas Liebig received funding from the European UnionHorizon 2020 Programme (Horizon2020/2014-2020), undergrant agreement number 688380 “VaVeL: Variety, Veracity,VaLue: Handling the Multiplicity of Urban Sensors”.

REFERENCES

[1] N. Bui, M. Cesana, S. A. Hosseini, Q. Liao, I. Malanchini, and J. Widmer,“A survey of anticipatory mobile networking: Context-based classifi-cation, prediction methodologies, and optimization techniques,” IEEECommunications Surveys & Tutorials, 2017.

[2] B. Sliwa, T. Liebig, R. Falkenberg, J. Pillmann, and C. Wietfeld,“Efficient machine-type communication using multi-metric context-awareness for cars used as mobile sensors in upcoming 5G networks,” toappear in IEEE Vehicular Technology Conference (VTC-Spring), Porto,Portugal, Jun 2018. [Online]. Available: https://arxiv.org/abs/1801.03290

[3] C. Ide, B. Dusza, and C. Wietfeld, “Client-based control of the in-terdependence between LTE MTC and human data traffic in vehicularenvironments,” IEEE Transactions on Vehicular Technology, vol. 64,no. 5, pp. 1856–1871, May 2015.

[4] B. Sliwa, “Raw experimental cellular network quality data,” Oct 2017.[Online]. Available: http://doi.org/10.5281/zenodo.1035079