Channel-Independent Viterbi Algorithm (CIVA) for DNA

Download Report

Transcript Channel-Independent Viterbi Algorithm (CIVA) for DNA

Power Efficient Wireless Sensor Networks with Distributed Transmission-Induced Space Spreading

Xiaohua (Edward) Li and N. Eva Wu Department of Electrical and Computer Engineering State University of New York at Binghamton {xli, evawu}@binghamton.edu

http://ucesp.ws.binghamton.edu/~xli 1

Major Contributions •

Resolve the conflict between transmission energy efficiency and fault tolerance

Propose distributed space-spreading for 1. Efficient/robust blind signal detection 2. Transmission energy efficiency 3. Network reliability 2

1.1. Sensor Network Challenges

• How to improve transmission energy efficiency in deep faded near-ground communications?

• How to improve fault tolerance and network reliability with low cost sensors suffering from high failure rate?

• How to resolve the conflict between energy efficiency and fault tolerance? They have contradictory requirements on redundancy.

Multi-hop Wireless Sensor Network 3

1.2 Strategies for the Challenges

• Distributed multi-transmission with space spreading – Transmission redundancy provides diversity for energy efficiency – Transmission redundancy provides fault tolerance Scrambled Parallel Transmission from J Sensors 4

• Why can we use multi-transmission?

– Wireless transmission is broadcasting • a data packet can be received/retransmitted by multiple sensors – There are always multiple standby sensors ready for multi-transmission • Energy in standby state is in the same level as in receiving state • How to perform multi-transmission?

– Distributed space-spreading : scrambled parallel transmission (the above figure) – Distributed space-time coding: to appear in

Electronics Letters

, 2003.

5

2. Space Spreading and Blind Symbol Estimation

• Sensor

j

in cluster

i

transmits the same signal

s

(

n

) with different PN scrambling:

s j

j

(

n

) 

s

(

n

)

c j

(

n

), 1 ,  ,

J

• A sensor in cluster

i+1

receives (separately) signals from all sensors

j

:

x

j

(

n

)  

j

H

j

s

j

(

n

) 

v

j

(

n

), 

j

: Rayleigh fading H

j

    

h j

( 0 )  

h j

(

L h

)

h j

( 0 )  

h j

(

L h

)     6

 Blind channel estimation from cross correlatio n :

R

ji

R

j

E

[

x

j

(

n

)

c

*

j

(

n

d

)

x

i H

(

n

)

c i

(

n

d

)],  [

R

j

1 ,  ,

R

j

,

j

 1 ,

R

j

,

j

 1 ,  ,

R

jJ

] because

R

ji

 

j

i

h

j

(

d

)

h

i H

(

d

) 

s

2 and

R

j

are rank 1.

 Blind equalizati on of signals from each sensor j : arg min ||

f

j H

x

j

(

n

) || 2 ,

s

.

t

.,

f

j H j

 1  Blind symbol estimation from diversity combining

s

ˆ (

n

d

) 

j J

  1

f

j H

x

j

(

n

)

c

*

j

(

n

d

) : 7

3. Energy Efficiency Analysis

• Transmission energy efficiency comes from the diversity of the scrambled parallel transmission • Major results: Propositio n 1.

Let the total power be

JE

[ 

j

2 ] 

s

2 , and the power of each sensor in multi transmissi on be

E

[ 

j

2 ] 

s

2 .

If

E

[ 

j

2 ] 

s

2 / 

v

2  1 , multi transmissi on has much less symbol error rate (SER), or can use much less power to achieve the same SER than single transmissi on.

8

Propostion For multi 2 .

Let transmissi  ~ 

E

[ 2 ] 

s

2 be the power of on, there exists 

J

 single  ~ such that transmissi on.

P

[

j J

  1  2

j

A

] 

P

[ 2 

A

].

Power ratio of single-transmission to multi-transmission for 15 dB SNR with Probability B.

Multi-transmission can be more than 30 dB more energy efficient.

9

4. Fault-Tolerance Analysis

 Goal : maximize signal availabili ty for the overall network, determine active sensor number J for each cluster  K hop network failure probabilit y :

F F

(

t

)  1  

i K

 1 [ 1 

F i F

(

t

)],

F i F

(

t

) 

k r

 1   0  

J i r

  [

R Snsr

(

t

)]

r

[ 1 

R Snsr

(

t

)]

J i

r

,

R Snsr

(

t

) 

e

 

t

 Problem : For each cluster, select the number of active sensors

J i

 [

J i

, min ,

J i

, max ], such that each cluster achieves the highest probabilit y at a specified network life

T D

 Solutions : Determine

J i

that has the smallest

F i F

(

t

).

Use constraine d dynamic programmin g for the smallest

F F

(

t

).

10

Cumulative Distribution Function of the i-th Sensor Cluster (  /I i =0.0001 failures/packet) 0.25

1 participating sensor/packet transmission 2 participating sensors/packet transmission 10 participating sensors/packet transmission 0.2

0.15

0.1

0.05

0 0 200 400 600 800 1000 1200 1400 1600 1800 Number of transmitted packets through the i-th sensor cluster 2000 • Example: For design life T D =2000 packets, J=2 is better and has reliability 0.89. However, for reliability 0.99, J=10 is better, though with a shorter design life T D =1000 packets.

11

5. Simulations

Multi-transmission : New batch & adaptive algorithms: J=8 sensors.

Single-transmission : DSSS with Rake receiver: processing gain 8.

Blind CMA Training MMSE equalization 12

Compare space-spreading with spectrum-spreading (DSSS) 13

Transmi ssion Power New Batch 1 New Adaptive Training MMSE 1.12

14.1

Blind CMA >89 DSSS/ Rake 7.1

• Transmission power (normalized with that of the new batch algorithm) required to achieve symbol-error-rate (SER) 0.01 • Multi-transmission-based space-spreading has higher energy efficiency, longer sensor lifetime, and higher reliability.

14

6. Conclusions • Propose a new space-spreading scheme for wireless sensor networks to achieve

– transmission energy efficiency – blind symbol estimation – transmission/network reliability • Resolve the conflict between energy efficiency and fault tolerance via transmission redundancy 15