Adaptive Decentralized Control of Underwater Sensor Networks for - - PowerPoint PPT Presentation

▶
adaptive decentralized control of underwater sensor
SMART_READER_LITE
LIVE PREVIEW

Adaptive Decentralized Control of Underwater Sensor Networks for - - PowerPoint PPT Presentation

Sep 20, 2022 443 likes •1.04k views

Adaptive Decentralized Control of Underwater Sensor Networks for Modeling Underwater Phenomena Carrick Detweiler , Marek Doniec , Mingshun Jiang , Mac Schwager , Robert Chen , Daniela Rus Massachusetts Institute of

slide-1
SLIDE 1

Adaptive Decentralized Control of Underwater Sensor Networks for Modeling Underwater Phenomena

Carrick Detweiler‡†, Marek Doniec†, Mingshun Jiang⋆, Mac Schwager†, Robert Chen⋆, Daniela Rus†

† Massachusetts Institute of Technology ‡ University of Nebraska-Lincoln ⋆ University of Massachusetts–Boston

SenSys 2010, Zurich, Switzerland

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 1

slide-2
SLIDE 2

Motivation: Underwater Sensing

BP oil spill – riser pipe

Image from reuters.com SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 2

slide-3
SLIDE 3

Motivation: Underwater Sensing

BP oil spill – extent is unknown

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 3

slide-4
SLIDE 4

Motivation: Underwater Sensing

Boston Harbor sewer pipe output

Image courtesy Mingshun Jiang, UMass Boston SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 4

slide-5
SLIDE 5

System Approach

Many inexpensive sensors Networked for real-time feedback Collaborate with robot Adjust depth for sensing using decentralized depth control algorithm (*this talk*) Depth adjustment enables: Easy deployment Easy recovery GPS or radio on surface Optimizing position for:

Sensing Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 5

slide-6
SLIDE 6

Related Work

Water column profilers

Glenn et al.. The leo-15 costal cabled observatory – phase II for the next evolutionary decade of oceanography. Scientific Submarine Cable, 2006. Howe and McGinnis. Sensor networks for cabled ocean observatories. International Symposium on Underwater Technology, 2004.

  • Joeris. A horizontal sampler for collection of water samples near the bottom. Limnology and Oceanography, 1964.

Coverage and sensor placement

Bullo, Cort´ es, and Mort´ ınez. Distributed Control of Robotic Networks. Applied Mathematics Series. Princeton University Press, 2009. Guestrin, Krause, and Singh. Near-optimal sensor placements in gaussian processes. International Conference on Machine Learning, 2005. Ko, Lee, and Queyranne. An exact algorithm for maximum entropy sampling. Operations Research, 1995. Schwager, Rus, Slotine. Decentralized, adaptive coverage control for networked robots. IJRR, 2009.

Simulated underwater depth adjustment algorithms

Akyildiz, Pompili, and Melodia. State-of-the-art in protocol research for underwater acoustic sensor networks. WUWNet, 2006. Cayirci, Tezcan, Dogan, and Coskun. Wireless sensor networks for underwater survelliance systems. Ad Hoc Networks, 2006.

Related areas

Drifting floats AUVs adjusting relative positions SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 6

slide-7
SLIDE 7

Outline

1

Motivation and Overview

2

Related Work

3

Approach

4

Decentralized Sensing Optimization Algorithm Simulation Results AquaNode Underwater Sensor Network Experimental Results

5

Future Work

6

Conclusions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 7

slide-8
SLIDE 8

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-9
SLIDE 9

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-10
SLIDE 10

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-11
SLIDE 11

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-12
SLIDE 12

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-13
SLIDE 13

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-14
SLIDE 14

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-15
SLIDE 15

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-16
SLIDE 16

Communication

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 8

slide-17
SLIDE 17

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-18
SLIDE 18

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-19
SLIDE 19

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-20
SLIDE 20

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-21
SLIDE 21

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-22
SLIDE 22

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-23
SLIDE 23

Improving Sensing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 9

slide-24
SLIDE 24

Contributions

Decentralized depth control algorithm Optmizes depths for sensing Based on covariance measurements Provable convergence Low processing and communication Tested in simulation Implemented and tested on AquaNodes

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 10

slide-25
SLIDE 25

Decentralized Depth Adjustment for Improved Sensing

Measurement of water column properties

Temperature, salinity, pH, dissolved O2, etc. Images

Capture time-varying properties Constraints

Power

Minimize motion Minimize communication

Acoustic communication bandwidth

11 bytes per packet

Transmit just position, depth, and sensor reading

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 11

slide-26
SLIDE 26

Covariance

Measure q1 from constrained path Changes at q1 correlated to changes at p1 Highest correlation when p1 close to q1: Min(Dist(q1, p1)) More generally use covariance: Max(Cov(q1, p1)) Allows different sensing functions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 12

slide-27
SLIDE 27

Covariance

Measure q1 from constrained path Changes at q1 correlated to changes at p1 Highest correlation when p1 close to q1: Min(Dist(q1, p1)) More generally use covariance: Max(Cov(q1, p1)) Allows different sensing functions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 12

slide-28
SLIDE 28

Covariance

Measure q1 from constrained path Changes at q1 correlated to changes at p1 Highest correlation when p1 close to q1: Min(Dist(q1, p1)) More generally use covariance: Max(Cov(q1, p1)) Allows different sensing functions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 12

slide-29
SLIDE 29

Covariance Model

Assume Gaussian Different variance along surface Better models with more knowledge

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 13

slide-30
SLIDE 30

Covariance Model

Assume Gaussian Different variance along surface Better models with more knowledge

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 13

slide-31
SLIDE 31

Multiple Points Problem

Idea: maximize sum over whole region:

  • Q

N

  • i=1

Cov(q, pi) Problem: left and right are same: .5 + .5 + .5 = 1.5 .25 + .25 + .5 + .5 = 1.5

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 14

slide-32
SLIDE 32

Multiple Points Solution

Solution: invert sum of covariance and minimize

  • Q

N

  • i=1

Cov(q, pi) −1 Yields:

1 .5+.5 + 1 .5 = 3 1 .5+.25 + 1 .5+.25 = 22 3

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 15

slide-33
SLIDE 33

Algorithm Approach

Objective function: H(p1, ..., pN) =

  • Q

N

  • i=1

Cov(q, pi) −1 dq Decentralized gradient controller: ˙ pi = −k ∂H ∂zi ∂H ∂zi =

  • Q

g(q, p1, ..., pN)2f (pi, q)(zi − zq) σ2

d

dq g(q, p1, ..., pN) = N

  • i=1

f (pi, q) −1 f (pi, q) = Cov(pi, q) = Ae

−

  • (xi −xq)2+(yi −yq)2

2σ2 s

+ (zi −zq)2

2σ2 d

  • SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu)

16

slide-34
SLIDE 34

Algorithm Convergence

Each node moves according to: ˙ pi = −k ∂H ∂zi Theorem: decentralized controller converges to local minimum Proof: convergence proof using Lyapunov criteria

H must be differentiable;

∂H ∂zi must be locally Lipschitz;

H must have a lower bound; H must be radially unbounded or the trajectories of the system must be bounded.

Verified in simulation, pool, and river experiments

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 17

slide-35
SLIDE 35

Simulation Results: Versus Matlab’s fminsearch

5 10 15 20 25 30 50 100 150 200 250

Total Search Time Number of Nodes Minutes fminsearch distributed controller

5 10 15 20 25 30 1500 2000 2500 3000 3500 4000 4500 5000 5500 6000

Objective Value Number of Nodes Obj Val (smaller better) fminsearch distributed controller

fminsearch: Matlab’s nonlinear unconstrained solver Much faster runtime Typically lower objective value

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 18

slide-36
SLIDE 36

Posterior Variance

50 100 150 200 250 300 350 400 0.78 0.79 0.8 0.81 0.82 0.83

posterior error time [sec] 00 min 10 sec 01 min 40 sec 03 min 20 sec 05 min 00 sec 06 min 40 sec

Posterior variance

Variance given sensor positions, assuming Gaussian process σ2

q|P = Cov(q, q) − Σq,P · Σ−1 P,P · ΣP,q

Requires matrix inversion (O(n2) memory for n sensors)

Decentralized depth control algorithm

Tends to reduce posterior error Constant memory requirements

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 19

slide-37
SLIDE 37

Simulation Results: Data Reconstruction

Top row: Original data Bottom row: Depth adjustment algorithm

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 20

slide-38
SLIDE 38

Outline

1

Motivation and Overview

2

Related Work

3

Approach

4

Decentralized Sensing Optimization Algorithm Simulation Results AquaNode Underwater Sensor Network Experimental Results

5

Future Work

6

Conclusions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 21

slide-39
SLIDE 39

Underwater Sensor Network: AquaNodes

Multi-purpose underwater sensor network Acoustic, optical, and radio communication Easy to use and deploy Dynamic depth adjustment

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 22

slide-40
SLIDE 40

Dynamic Depth Adjustment

assembled winch timing belt motor spool with fishing line glass thrust bearing transducer magentic couplers

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 23

slide-41
SLIDE 41

Dynamic Depth Adjustment

anchor line delrin pulley wheel bronze bushing spool bronze bushing aluminum shaft magnetic coupler magnetic coupler timing belt spur gearhead motor

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 24

slide-42
SLIDE 42

Dynamic Depth Adjustment

Video: Winch in Pool

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 25

slide-43
SLIDE 43

AquaNodes: Platform Overview

LPC2148 60MHz ARM7 SD Card for logging Temperature, pressure, CDOM, salinity, dissolved 02, camera Digital and analog inputs Depth adjustment: 2.4m/min Communications

Acoustic (FSK modulation): 300b/s up to 200m Radio (1W 900MHz Aerocomm): 57kb/s up to 1km on surface Optical (DPIM modulation): 3Mb/s up to 5m

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 26

slide-44
SLIDE 44

Decentralized Depth Adjustment Results

Control Output

500 1000 1500 2000 10

−2

10 10

2

10

4

10

6

10

8

Hdz time [sec]

Experiment 2

Depths

500 1000 1500 2000 −3 −2.5 −2 −1.5 −1 −0.5

depth [m] time [sec]

Node 1 Node 2 Node 3 Node 4

Four AquaNodes running depth control algorithm in pool Three iterations of depth control algorithm Algorithm converges within 10 minutes Nodes spread out

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 27

slide-45
SLIDE 45

Decentralized Depth Adjustment Communication

Number of Neighbor

50 100 150 200 250 300 350 400 450 500 1 2 3 Node 1 valid node count time [sec] 1 2 3 Node 2 valid node count 1 2 3 Node 3 valid node count 1 2 3 Node 4 valid node count

Depth

50 100 150 200 250 300 350 400 −25 −20 −15 −10 −5 Node 1 depth [m] time [sec] −25 −20 −15 −10 −5 Node 2 depth [m] −25 −20 −15 −10 −5 Node 3 depth [m] −25 −20 −15 −10 −5 Node 4 depth [m]

Communication data from part of previous experiment (4 nodes) Nodes do not hear all other nodes Algorithm handles communication dropouts

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 28

slide-46
SLIDE 46

Decentralized Depth Adjustment Communication

50 100 150 200 250 300 350 400 −25 −20 −15 −10 −5

Node 1 depth [m] time [sec]

−25 −20 −15 −10 −5

Node 2 depth [m]

−25 −20 −15 −10 −5

Node 3 depth [m]

−25 −20 −15 −10 −5

Node 4 depth [m] SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 29

slide-47
SLIDE 47

Changing Covariance

6400 6600 6800 7000 7200 7400 7600 7800 8000 50 100 150 200 250 300 350 400 Time (s) Obj Value and Depth (cm) Node 0 Node 1 Node 2 Node 3 River Depth (cm)

Changing covariance over time For example tidal changes Objective value returns to minimum after algorithm adjusts

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 30

slide-48
SLIDE 48

Outline

1

Motivation and Overview

2

Related Work

3

Approach

4

Decentralized Sensing Optimization Algorithm Simulation Results AquaNode Underwater Sensor Network Experimental Results

5

Future Work

6

Conclusions

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 31

slide-49
SLIDE 49

Future Work

Collect scientific data Long-term deployments Determine maximum water current Examine impact of bio-fouling Leverage depth adjustment for other applications

Optimize Acoustic Communication Multi-modal communication (acoustic, radio, optical)

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 32

slide-50
SLIDE 50

Neponset River Experiment

Summer deployment in Neponset River w/ 4 nodes Nodes performed column scans, sensing temp, pressure, CDOM Collecting data for future depth optimization experiments

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 33

slide-51
SLIDE 51

Neponset River Experiment

1 2 3 4 5 6 2 4

Depth (m)

Node 2 1 2 3 4 5 6 2 4 6

Line Length (m)

1 2 3 4 5 6 2 4

CDOM (Volt)

1 2 3 4 5 6 20 25 30

Temp (C) Experiment Hour

Deployment for half tidal cycle

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 34

slide-52
SLIDE 52

Acoustic Communication Example

0.0 m 0.5 m 1.0 m 1.5 m 2.0 m 0.0 m 0.5 m 1.0 m 1.5 m 2.0 m

node A depth node B depth accoustic modem success rate

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

Placement is critical for acoustic comms Short-range river experiment between walls

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 35

slide-53
SLIDE 53

Contributions and Conclusions

Algorithms in an underwater sensor network

Decentralized depth control for sensing

Provable convergence

Verified in simulation and field experiments

System implementation and experiments

Underwater sensor network Dynamic depth adjustment Tested in pools, lakes, and rivers

Future work taking advantage of depth adjustment Leverage sensor networks to improve environmental understanding

contact me at: carrick@cse.unl.edu http://cse.unl.edu/∼carrick

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 36

slide-54
SLIDE 54

Questions?

Algorithms in an underwater sensor network

Decentralized depth control for sensing

Provable convergence

Verified in simulation and field experiments

System implementation and experiments

Underwater sensor network Dynamic depth adjustment Tested in pools, lakes, and rivers

Future work taking advantage of depth adjustment Leverage sensor networks to improve environmental understanding

contact me at: carrick@cse.unl.edu http://cse.unl.edu/∼carrick

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 37

slide-55
SLIDE 55

Outline All

1

Motivation and Overview

2

Related Work

3

Approach

4

Decentralized Sensing Optimization Algorithm Simulation Results AquaNode Underwater Sensor Network Experimental Results

5

Future Work

6

Conclusions SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 38

slide-56
SLIDE 56

Radio Communication

Uses winch to go to surface 900MHz Aerocomm radio Built-in broadcast protocol 1 Watt transmit power 20km max range 1km typical range

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 39

slide-57
SLIDE 57

Acoustic Communication

Developed in our lab Broadcast protocol 600MHz DSP 27-33 KHz Frequency-Shift Keying (FSK) 300b/s 45mJ/bit (2W transmit power) 400m range Ranging between modems

4cm resolution

Time Division Multiple Access (TDMA)

Self-synchronizing

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 40

slide-58
SLIDE 58

Optical Communication

Developed in our lab Point-to-Point 5 meter 90◦ cone 3Mbit/s 7µJ/bit 532nm wavelength (green) Digital Pulse Interval Modulation (DPIM) modulation

SenSys 2010–Carrick Detweiler (carrick@cse.unl.edu) 41