algorithms for optimal power allocation of wireless multi ... · industrial applications using the...

Algorithms for Optimal Power Allocationof Wireless Multi-Hop Heterogeneous Networks

under Statistical Delay Constraints

Neda Petreska(joint work with J. Gross and H. Al-Zubaidy)

Fraunhofer Institute for Embedded Systems andCommunication Technologies ESK

WoNeCa 2018

Erlangen, 28.02.2018

N. Petreska (Fraunhofer ESK) WoNeCa 2018 Erlangen, 28.02.2018 1 / 17

Wireless Industrial Sensor Networks


Wireless Industrial Sensor Networks

Provide e2e delay guarantees

Maximize battery lifetime

Minimize interference


System Model

Multi-hop path

Time slotted system

Block-fading channels with non-identicallydistributed, but statistically independentchannel gains

a b


Open Questions

How to analytically define the end-to-end performance guarantees forwireless industrial networks?⇒ Consider fading and queuing effects.

Is an optimal transmit power allocation possible?

Does the analytical optimum resemble the real system optimum?


Used Method: Stochastic (min,x) Network Calculus

Bit domain

SNR domain


Analytical Delay Bound


Recursive End-to-End Delay Bound

L = {1, 2, 3}

A(t) D(t)

S1 S2 S3

Violation probability of the target end-to-end delay w :

K{1,2,3}(w) =MS2

MS2 −MS3

· K{1,2}(w) +MS3

MS3 −MS2

· K{1,3}(w)

N.Petreska, H.Al-Zubaidy, R.Knorr, J.Gross, ”On the Recursive Nature of End-to-End Delay Bound for Heterogeneous WirelessNetworks”, ICC, 2015


Transmit Power Minimization Algorithm

Enable delay-aware dynamic power management to

Extend battery ⇒ node ⇒ network lifetime

Reduce interference

Enable coexistence of several wireless technologies

Use delay bound convexity


Convex Delay Bound: One Hop

0 0.005 0.01 0.015 0.02 0.025 0.03 0.035 0.04

s

10-8

10-6

10-4

10-2

100

Del

ay V

iola

tion

Pro

babi

lity

SNR=[5] dBSNR=[7] dBSNR=[10] dBSNR=[12] dB


Convex Delay Bound: Multi-Hop

s0 0.005 0.01 0.015 0.02 0.025 0.03 0.035

Del

ay V

iola

tion

Pro

babi

lity

10-7

10-6

10-5

10-4

10-3

10-2

10-1

100

SNR=[2, 4, 6] dBSNR=[5, 7, 9] dBSNR=[8, 10, 12] dBSNR=[11, 13, 15] dB

N.Petreska, ”End-to-End Performance Analysis for Industrial IEEE 802.15.4e-based Networks”, Fachgesprach fur Sensornetze,2017


Binary Search Algorithm

N.Petreska, H.Al-Zubaidy and J.Gross, ”Power Minimization for Industrial Wireless Networks Under Statistical DelayConstraints”, ITC, 2014


Power Minimization Algorithm


Link Heterogeneity

Delay in superframes0 2 4 6 8 10 12 14 16 18 20

Net

wor

k lif

etim

e ex

tens

ion

in %

0

5

10

15

20

25

d=[20, 19, 21], Norm=4d=[20, 30, 10], Norm=40d=[5, 28, 27], Norm=46d=[20, 35, 5], Norm=60d=[5, 40, 15], Norm=70d=[5, 50.5, 4.5], Norm=92

Target delay violation probabilityε = 10−3.

Payload size 10 B.

Transmit power used as acomparison Ptx = 4 dBm.

Duration of superframe 30 ms.


Bound-Based vs. Real System Optimum

How well the NC-based power optimization reflects the real systemoptimum?

Delay in superframes0 5 10 15

Min

imal

Bat

tery

Dur

atio

n [#

supe

rfra

mes

] #107

2.6

2.7

2.8

2.9

3

3.1

3.2

3.3

Norm 4 Alg.Norm 4 Sim.Norm 46 Alg.Norm 46 Sim.Norm 60 Alg.Norm 60 Sim.Norm 70 Alg.Norm 70 Sim.


Additional Network Lifetime Extension

Delay in superframes0 5 10 15

Add

ition

al N

etw

ork

Life

time

Ext

ensi

on [%

]

0

1

2

3

4

5

6

7

Norm 4Norm 46Norm 60Norm 70


Conclusions

Latency, reliability and energy efficiency - crucial requirements forindustrial applications

Using the (min,x) network calculus we provide

a closed form expression for the end-to-end delay bound in multi-hopwireless heterogeneous networksa bound-based optimal power allocation algorithm

Bound-based power allocation can be used to design reliable andpower-efficient wireless industrial networks


Next Steps

Validate the delay bound in real IEEE 802.15.4e testbed

Currently working with ContikiNG and Cooja simulation networkBuild a multi-hop network prototype and test the power savings undervarious delay and reliability constraints

Use the recursive behaviour of the end-to-end delay bound to define apower-efficient routing algorithm

Zolertia motesTI CC2538System-On-Chip


Extra Slides


Validation of the WirelessHART Service Curve

Delay [ms]0 50 100 150 200 250 300 350 400 450 500

Del

ay V

iola

tion

Pro

babi

lity

10-5

10-4

10-3

10-2

10-1

100

SNR=[4] dBSNR=[4, 6] dBSNR=[4, 6, 8] dBSNR=[4, 6, 8, 5] dBSNR=[4] dB Sim.SNR=[4, 6] dB Sim.SNR=[4, 6, 8] dB Sim.SNR=[4, 6, 8, 5] dB Sim.

N.Petreska, H.Al-Zubaidy, B. Staehle, R. Knorr and J.Gross, ”Statistical Delay Bound for WirelessHART Networks”,PE-WASUN, 2016


algorithms for optimal power allocation of wireless multi ... · industrial applications using the...

Documents