approaches to improving acoustic communications …gobysoft.org/dl/ucomms_2012_talk.pdfgobysoft.org...

Schneider & Schmidt: Improving Acomms on AUVsUComms 2012 - Sestri Levante, Italy

1

Toby SchneiderMIT/WHOI Joint Program

Henrik SchmidtMIT Laboratory for Autonomous Marine Sensing Systems

Approaches to Improving Acoustic Communications on

Autonomous Mobile Marine Platforms

...01010100101010100101010001...

gobysoft.orgSchneider & Schmidt: Improving Acomms on AUVsUComms 2012 - Sestri Levante, Italy

2

1. Autonomous underwater vehicles (AUVs) need wireless data transfer:

Collaborative missions•

Collected data•

Commands•

2. Current state-of-the-art information throughput is un-acceptable

3. However, AUVs have:

Mobility•

“Intelligence” - modeling, state awareness•

Overview


3

Can we use the AUV’s “intelligence” to increase acoustic comms throughput (useful information / unit energy)?

Two (complementary) approaches:

Disruptive• (affects vehicle course)- stop to transmit or receive (reduce Doppler or self-noise)- move to expected beneficial position (SNR maximum)

Non-disruptive• (transparent to mission)- frequency band or transmit power selection- mission-based source coding

Overview


4

Use measured environment to generate acoustic model to drive adaptation in depth (assume range is out of our control). Goal: improve SNR from buoy to AUV

GLINT10

BTech Acoustics, LLC Acoustic Transducer and Sonar Development Company

17 Surrey Road, Barrington, RI 02806 Phone: (401) 261-9318 E-mail: [email protected]

BTech’s Model BT-2RCL transducer is a proven standard for underwater acoustic communica-tion (ACOMMS). Specifications (Nominal):

Resonance Frequency (fr): 28 kHz Suggested Range: 20 – 40 kHz TVR at fr: 141 dB re 1 Pa/V OCVS at fr: -191 dB re 1 V/Pa Horizontal Beam Pattern: Omnidirectional Vertical Beam Pattern: Toroidal

3.3751.500

10-32

1.875

Ø 1.875

Pin 1 (-), Pin 3 (+)

Performance Curves (Nominal):

101520253035404550110

115

120

125

130

135

140

145

150Transmitting Voltage Response (TVR)

Frequency [kHz]

TVR [dB re 1 µPa / V @ 1 m]

101520253035404550−220

−215

−210

−205

−200

−195

−190

−185

−180Open Circuit Voltage Sensitivity (OCVS)

Frequency [kHz]

OCVS [dB re 1 V / µPa @ 1 m]

10152025303540455010

11

12

13

14

15

16

17

18Admittance (Cp & G) in Water

Frequency [kHz]

Parallel Capacitance [nF]

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Conductance [mS]

0

30

6090

120

150

180

210

240270

300

330

0

−10

−20

−30

−40

−30

−20

−10

0 dB Scale

−3

20 kHz25 kHz30 kHz

©2010 BTech Acoustics, LLC. Specifications subject to change without notice. Rev. 02/10

Vertical BP

Model BT-2RCL

BTech Acoustics, LLC Acoustic Transducer and Sonar Development Company

17 Surrey Road, Barrington, RI 02806 Phone: (401) 261-9318 E-mail: [email protected]

BTech’s Model BT-2RCL transducer is a proven standard for underwater acoustic communica-tion (ACOMMS). Specifications (Nominal):

Resonance Frequency (fr): 28 kHz Suggested Range: 20 – 40 kHz TVR at fr: 141 dB re 1 Pa/V OCVS at fr: -191 dB re 1 V/Pa Horizontal Beam Pattern: Omnidirectional Vertical Beam Pattern: Toroidal

3.3751.500

10-32

1.875

Ø 1.875

Pin 1 (-), Pin 3 (+)

Performance Curves (Nominal):

101520253035404550110

115

120

125

130

135

140

145

150Transmitting Voltage Response (TVR)

Frequency [kHz]

TVR [dB re 1 µPa / V @ 1 m]

101520253035404550−220

−215

−210

−205

−200

−195

−190

−185

−180Open Circuit Voltage Sensitivity (OCVS)

Frequency [kHz]

OCVS [dB re 1 V / µPa @ 1 m]

10152025303540455010

11

12

13

14

15

16

17

18Admittance (Cp & G) in Water

Frequency [kHz]

Parallel Capacitance [nF]

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Conductance [mS]

0

30

6090

120

150

180

210

240270

300

330

0

−10

−20

−30

−40

−30

−20

−10

0 dB Scale

−3

20 kHz25 kHz30 kHz

©2010 BTech Acoustics, LLC. Specifications subject to change without notice. Rev. 02/10

Vertical BP

Model BT-2RCL

30 m

110

m

r

dWHOI Micro-Modem (25 kHz, FH-FSK)

Buoy

AUV


5

AUV meaures SSP using CTD via • “yoyo” depth excursions.

Thermocline-adaptive sampling• (Petillo 2010) used.

1. Measure Environment

modeled acoustics

generic acoustic model

modeled communications

actions

application-specific modeling

AI reasoning

» communications quality» acoustic pressure» environment (S,T,P)

measured

explore physical space &data from other agents

modeled beamforming

environmental data

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

Temperature

Dept

hThermocline}

S. Petillo: Adapted with permission


6

2. Model Acoustics (GRAM + BELLHOP)

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental dataenvironmental data


actions


AI reasoning


measured


modeled beamforming

1500 1520 1540

0102030405060708090

100110 0 500 1000 1500 2000

0102030405060708090

100110

2025303540455055606570

dept

h (m

)

range (m)sound speed (m/s)

tran

smiss

ion

loss

(dB)

10 20 30

0

10

20

30

40

50

60 10 20 30

0

10

20

30

40

50

60 15 20 25

0

10

20

30

40

50

60 10 20 30

0

10

20

30

40

50

60 10 20 30

0

10

20

30

40

50

60

dept

h (m

)

RMS delay spread (ms)

r = [0,400) m r = [400,800) m r = [800,1200) m r = [1200,1600) m r = [1600,2000) m

SSP TL

Delay Spread


7

Averaged modeled acoustic intensity in range for future horizon:

based on AUV’s instantaneous speed,• heading

scaled by local extrema or • overall utility metric from past data

3. BHV_AcommsDepth Utility Function

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental data

modeled acoustics


environmental data

actions

AI reasoning


measured


modeled beamforming

60 80 100

0

20

40

60

80

100

120

r0 = 0 m

44 46 48

0

20

40

60

80

100

120

r0 = 800 m

30 40 50

0

20

40

60

80

100

120

r0 = 1600 m

dept

h (m

)

utility (%)

0% utility: TL = 108 dB (<1% messages received)

100% utility: TL = 0 dB


8

Interaction with other behaviors:

3. BHV_AcommsDepth Utility Function

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental data

modeled acoustics


environmental data

actions

AI reasoning


measured


modeled beamforming

0 20 40 60 80 100

0

20

40

60

80

100

120

BHV_AcommsDepthDepth Avoidance (40m)

Minimum Altitude (20m)Mean

dept

h (m

)

utility (%)


9

2010-08-08 runs:

white dots: AUV position•

Buoy: range = 0 m, depth = 30m•

Underlay: TL model from • representative SSP.

4. Transit

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental data

modeled acoustics




modeled beamforming

environmental data


measured


0 200 400 600 800 1000 1200 1400 1600 1800 2000

0

20

40

60

80

100

Range (m)

Dept

h (m

)

20

30

40

50

60

70


10

5. Measure data and feedback

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental data

modeled acoustics



actions


AI reasoning

modeled beamforming

environmental data

Measured data:

1. Vehicle depth: - random variable D

2. Received 32 byte messages (good = passed CRC without errors):

- Bernoulli random variable R (good / bad)

0 0.5 1

0

10

20

30

40

50

60

dept

h (m

)

dept

h (m

)

P(R = good | D = depth)P(D)P(D | R = good)

0 0.01 0.02

0

10

20

30

40

50

60

Bayes

r = [1600, 2000) m r = [1600, 2000) m


11

5. Measure data and feedback

modeled acoustics



actions


AI reasoning


measured


modeled beamforming

environmental data

modeled acoustics



actions


AI reasoning

modeled beamforming

environmental data

30 35 40

0

10

20

30

40

50

60

r = [0,400) m

45 50 55

0

10

20

30

40

50

60

r = [400,800) m

50 55 60

0

10

20

30

40

50

60

r = [800,1200) m

50 55 60

0

10

20

30

40

50

60

r = [1200,1600) m

55 60 65

0

10

20

30

40

50

60

r = [1600,2000) m

dept

h (m

)

TL (dB)

μ, σ for modeled TL for 111 measured SSPs

0 0.5 1

0

10

20

30

40

50

600 0.5 1

0

10

20

30

40

50

600 0.5 1

0

10

20

30

40

50

600 0.5 1

0

10

20

30

40

50

600 0.5 1

0

10

20

30

40

50

60

dept

h (m

)

P(R = good | D = depth)

Measured Bayesian probability of message receipt for given depth


12

Use mission knowledge and models to greatly improve compression of vehicle data, e.g. position in space (x,y,z).

Non-disruptive: source coding

Mission pathActual pathSent pointSent difference

y

x

(dx0,dy0)

datum(0,0)

Pdx

dxdx0

0 10

0-2-dxmin

-1-2 1 2 ......

dx0

dxmindx:

symbol:

dxmax

-1-dxmin

-dxmin1-dxmin

2-dxmin dxmax-dxminOOREOF


13

Position of single AUV doing many different missions over 60+ hours dive time:

Source coding: GLINT10 data

20004000

60008000

10000

2000

4000

6000

2010−08−01

2010−08−02

2010−08−03

2010−08−04

2010−08−06

2010−08−07

2010−08−08

2010−08−09

2010−08−10

x (m)y (m)

date

(UTC

)

2010−08−01 2010−08−02 2010−08−03 2010−08−04 2010−08−06 2010−08−07 2010−08−08 2010−08−09 2010−08−10

5

10

25

35

45

55

65

75

85

95

105

date (UTC)de

pth

(m)

2010−08−01 2010−08−02 2010−08−03 2010−08−04 2010−08−06 2010−08−07 2010−08−08 2010−08−09 2010−08−10

5

10

25

35

45

55

65

75

85

95

105

date (UTC)

dept

h (m

)

2010−08−01 2010−08−02 2010−08−03 2010−08−04 2010−08−06 2010−08−07 2010−08−08 2010−08−09 2010−08−10

5

10

25

35

45

55

65

75

85

95

105

date (UTC)

dept

h (m

)2010−08−01 2010−08−02 2010−08−03 2010−08−04 2010−08−06 2010−08−07 2010−08−08 2010−08−09 2010−08−10

5

10

25

35

45

55

65

75

85

95

105

date (UTC)

dept

h (m

)


14

Three models (s = AUV speed, τ = time between updates):

Adaptive: generated from cumulative frequencies1.

Gaussian (σ = sτ)2.

Uniform over [-5sτ, 5sτ)3.

Source coding: models

0

0.2

0.4

0

0.02

0.04

P(dx

)

−60 −40 −20 0 20 40 600

0.005

0.01

dx (meters)

adaptive

gaussian

uniform

2σ=2sτ


15

Encoded bits/message:

Adaptive: μ = 8.66, σ = 2.21.

Gaussian: μ = 14.46, σ = 0.162.

Uniform: μ = 18.16 , σ = 0.0803.

REMUS CCL: μ = 61, σ = 04.

int32 (uncompressed): μ = 96, σ = 05.

Source coding: results

4 5 6 7 8 9 10 12 14 16 18 20 30 40 50 60 70 80 9610 0

10 1

10 2

10 3

10 4

bits

coun

ts

adaptive

gaussian

uniform

int32


16

Two complementary techniques, which are in turn complementary to signal processing advances:

Disruptive• (changes vehicle motion)

Non-disruptive• (does not change mission)

As acomms become standardized, we want to see:

quantification of various physical effects on perfor-• mance. How much does multipath, SNR, Doppler mat-ter?

flexibility in choosing bit-rates, bandwidth, etc. to suite • vehicle mission needs & knowledge.

Summary


17

NATO Undersea Research Centre (La Spezia, Italy): • GLINT10 operations

Michael Porter: Acoustics Toolbox•

Michael Benjamin / Paul Newman: MOOS-IvP•

Lee Freitag & WHOI Micro-Modem group•

Contributors to the Open Source Software used in • GRAM / Goby

Acknowledgments

approaches to improving acoustic communications …gobysoft.org/dl/ucomms_2012_talk.pdfgobysoft.org...

Documents