mit marine robotics autonomous underwater vehicles

2/23/21

1

2.680 Spring 2021 – Marine Autonomy, Sensing and Communications – Applications

Web: http://oceanai.mit.edu/2.680

Email: Mike Benjamin, [email protected] Schmidt, [email protected]

2.680Unmanned Marine Vehicle Autonomy,

Sensing, and Communications

Lecture 2: Applications and Lessons Learned

February 23rd 2021


1


MIT Marine RoboticsAutonomous Underwater Vehicles

MIT Odyssey II (1995) Bluefin21 (2002, 2005)

Applications of Autonomous Underwater Vehicles:

• Mine countermeasures in shallow water

• Large-area Undersea Search and Surveilllance

• Deep-ocean oil exploration

• Autonomous Ocean Observation and Sensing Systems

• Under-ice environmental surveysBluefin Sandshark (2015)

2

2/23/21

2


Mine Counter Measuresin the Littoral Ocean

• Proud and buried mines• Detection - POD/PFA

• Localization - Navigation

• Classification POC/PFA• High Area Coverage rate

Objectives

Constraints

• Eliminate Divers• Autonomy

• Clandestine operation

3


Adaptive

Behavior

Cooperative

Behavior

Sonars

Uncertain Communication

Self-navigating

Network

Unknown Environment

No Maps

Mine Countermeasures withAutonomous Vehicle Networks

4

2/23/21

3


Sensing in the Marine Environment

Point Measurements

• Physical

– Temperature

– Pressure

• Chemical

– Salinity

– Natural plumes

– Pollutants

• Biological

– Plankton samples

Remote Sensing

• Vision

– High Resolution

– Short range (0-20 m)

– Examples

• Photography

• Fluorescence

• Optical backscatter

• Sonar

– Low-resolution

– Long range (0-5000km)

– Examples

• Side Scan Sonar

• Multi-beam Sonar

5


Seabed Mapping

Photo

Mosaic

675 kHz

Pencil-BeamSonar

Images Courtesy of H. Singh, WHOI

6

2/23/21

4


Side Scan Sonar Imaging

Aspect 270o Aspect 000o

Frequency 200-800 kHz• High range resolution

Wide horizontal aperture • Narrow horizontal beam• High angular resolution

Range

Angle

7


Dolphin Sonar -- Beampatterns

Figures from “The Sonar of Dolphins” by W. Au (Springer Verlag, 1993)

Wide beam -> Low angle resolution

8

2/23/21

5


Dolphin Sonar -- transmitted signals


Wide frequency band=>

Accurate ranging

9


Dolphin SonarReflection from Objects


10

2/23/21

6


GOATS’98

Odyssey II Bi-static Receiver Platform

11


GOATS’98 Experiment

Automated, Bistatic SAS Imaging

Super-critical Insonification

12

2/23/21

7


Navigating in the Ocean

• No GPS

• Optics, Radar or LORAN

• Only for ranges < 10-100 m

• Acoustic

• Long Baseline Navigation (LBL)

• Short and Ultra-short Baseline

Navigation (USBL)

• Simultaneous Localization and

Mapping (SLAM)

13


The Sonar of Bats


Wide beam -> Range only

14

2/23/21

8


CML - SLAMConcurrent Mapping and Localization

a) b)

c)SLAM:

Simultaneous Localization

And Mapping

15


MIT AUV OperationsBP’02 – MASAI’02

18

2/23/21

9


MIT-SACLANTCEN SAS Sonar

Source and Acquisition Payload Section

2x8-element Linear Array (7.5kHz)

16-element Linear Array (15kHz)

19


Target TrackingWide Beam Sonar

d

Horizontal

View

Vertical

View

Acceptable

Target Range

Time

20

2/23/21

10


BP’02 - MASAI’02SAS Zamboni Surveys

Navigation Sensors

•GPS (surface)

•Sonardyne LBL•DVL

•Compass

SAS Sonar

•4-16 kHz SBP Source

•2x8 element nose array

Klein 5000 (GESMA)

21


BP’02 – MASAI’02Simultaneous Localization and Mapping

22

2/23/21

11


Bat SonarSearch, Detection, and Tracking of prey

Figures from “The Sonar of

Dolphins” by W. Au (Springer Verlag, 1993)

2

3

4

1

5

4

3

2

1

5

Narrow band – Doppler Detection

Wide band – Range Tracking

23


Range-Doppler Resolution of Matched Filter

Ambiguity function of

CW sonar pulse signal


LFM sonar pulse signal


HFM sonar pulse signal

24

2/23/21

12


Detection Enhancement Using Adaptive Platform Control

Detectionmade

Adaptpath

Simulated Acoustic Data Signal to Reverb (SRR)

Adaptive

Planned

25


Environmentally Adaptive Sensing,

Communication and Autonomy

26

2/23/21

13


Acoustic Signaturesof Arctic Climate Change

• Sound speed profile

– Increased average sound speed

– Increased surface sound speed in open water creates efficient sound channel with

reduced surface interaction

– Recently observed warm water entering through Bering Strait at 100 m depth

creates very efficient duct without interaction with ice cover (‘Beaufort Lens’)

– Complex laterally inhomogeneous propagation environment in MIZ

• Ice cover

– Retreating ice cover

• Exposes environment to atmospheric interactions -> more temporal variability of acoustic environment

– Thinner ice with altered roughness statistics

• Changes in scattering loss for long-range propagation

• Changes in modal composition of long range propagation

– Changes in dominant ice fracturing processes

• More frequent, ice-mechanical events, e.g. ridging – 1D features in azimuth and range

• Less dominance of distributed floe boundary grinding – 2D azimuthally isotropic, homogeneous in range

– MIZ ambient noise characteristics becoming significant throughout Arctic.

27


WHOI ITP ProgramBeaufort Lens

MIT Laboratory forAutonomous Marine

Sensing Systems

ITP Locations Double DuctProfiles

ITP Buoy

ITP 84

28

2/23/21

14


The New Arctic Acoustic Environment

ITP 14Traditional Arctic SVP.

Monotonically increasing sound speed forces all paths to interact with ice

cover.

ITP 84New Arctic SVP

‘Beaufort Lens’

Dramatically improved

propagation conditions above 300 Hz in duct

isolated from surface and bottom interaction

New Arctic SSP, “Beaufort Lens” (ITP 84)

Range (km)

Tra

nsm

issio

n L

oss (

dB

)

10 20 30 40 50 60 70 80 90 100

Depth

(m

)

0

1000

2000

30001420 1460 1500

Sound Speed (m/s)

500

1500

2500

Traditional Arctic SSP (ITP 14)

Tra

nsm

issio

n L

oss (

dB

)

10 20 30 40 50 60 70 80 90 100

Depth

(m

)

0

1000

2000

30001420 1460 1500

500

1500

2500

Sound Speed (m/s) Range (km)

29


Beaufort Sea Acoustics Ice Camps

ICEX’16March 2016

AUV w. 32-element arrayVLA and towed HLA configurations

SIMI’94 –TAP’94March 1994

32-element VLA32 Element Mill’s Cross Array

30

2/23/21

15


ICEX16 Objectives• Operations

– Demonstrate ability to deploy and recover AUV with towed array from

14’x3’ hole in Arctic ice cover.

– Demonstrate ability to navigate AUV using INS with updates from acoustic tracking range shared with manned submarines.

– Communication through acoustic communication infrastructure shared with

manned submarines

– Demonstrate autonomous operations of several hours for collecting

scientific and tactical data from towed array, CTD, and upward looking

DVL.

• Scientific Data

– Ambient noise with vertical, horizontal and hybrid apertures for

characterizing directional properties of the ambient noise field in the new

Arctic.

– Spatial characterization of acoustic propagation in the new Arctic

– Ice roughness statistics using upward-looking DVL

– Under ice imaging using GoPro camera on AUV.

– CTD during all missions.

31


Arctic MIZ Reality

32

2/23/21

16


Macrura Survey MissionMarch 15, 2016

33


ICEX’16Integrated Acoustic Navigation and Communication

Acoustic Tracking and Navigation

Integrated with existing ARL/UW submarine tracking

WHOI HF Micro-Modem on platform emits tracking pulse

1PPS 3.5 ms CW at 13.5kHz with doublet every 10 seconds

10.5Khz carrier, 3kHz bandwidth 10ms FM sweep (“platform”) every 30 seconds

Topside tracker for aggressive outlier rejection, before fixes are transmitted back UUV via acoustic communication to update onboard INS navigation solution

Upward-looking DVL fusion disabled due to ice motion of order .5 kn

Acoustic CommunicationHardware: WHOI MF Micro-Modem (3.5 kHz center, 1.25 kHz bandwidth), shared with NUWC Digital Acomms transmit transducer and receive hydrophone

Software: Goby/DCCL marshalling, queuing, medium access, and physical layer interface.

DVL

INS

34

2/23/21

17


Untethered Survey MissionMarch 15, 2016

MIT Laboratory forAutonomous Marine Sensing Systems

35


Tracking Range Performance• Tracking range performance inferior to previous ICEX’

– Tracking range aperture smaller than historical due to ice floe

constraints

– Depth of tracking hydrophones fixed at 33 m

– Tracking of shallow targets (<80 m) have increased uncertainty beyond a ~ 1 km range and no tracking beyond ~ 2 km

– No tracking of deep targets (>80 m) beyond 1-1.5 km range

– Sporadic tracking at ranges 6-7 km

– Modeling confirms the performance degradation associated with

Beaufort Lens.

• Tracking range performance constrained AUV operations

– Safe operations restricted to area within range aperture (~ 1 km)

– Outlier rejection required on topside to avoid wasting ½ min update

slot

– 1 sec period ‘unit’ tracking pulse performed significantly better than

20 sec period ‘platform’ pulse

36

2/23/21

18


Acomms and Navigation in the Beaufort Sea Then and Now

RAP CorridorsHistorical Beaufort Sea

Shallow source/receiver provides direct paths to target at all operational

depths out to 6 km range. No reliance on slower

surface duct multipaths.

Present Beaufort Sea

Shallow source/receiver provides NO direct paths to shallow target.

Ice interaction degrades coherence beyond 0.5-1 km

range.Deeper CZ RAP paths at 6-7 km range.

Deep target reachable only at very short range.

37


pHelmIvPPlatform Autonomy

Acquisition

ArrayProcessing

Detection

Tracking

Classification

Localization

Data BusMOOSDB

pAcommsHandler

iFrontSeat

TDA

EmbeddedVirtual Ocean

pLamssMissionManagerMission Autonomy


Acquisition

ArrayProcessing

Detection

Tracking

Classification

Localization

Data BusMOOSDB

pAcommsHandler

iFrontSeat

TDA



Virtual Ocean Autonomy Testbed

‘

UUV Frontseat

Towed ArrayModel

µModem

Dynamic Model

pTDA

Topside C2

uSimINS uSimIDVL

pAcommsHandler

EmbeddedVirtual

Ocean

MOOS-IvP Payload Autonomy System


Acquisition

ArrayProcessing

Detection

Tracking

Classification

Localization

Data BusMOOSDB

pAcommsHandler

iFrontSeat

pTDA



GobyDCCL

‘Live’ Virtual Ocean

µModem

A/D-D/AA/D-D/A

38

2/23/21

19


ICEX2020 UUV Operations

Integrated Acoustic Tracking, Navigation and Communication

To ide

AUV

Ice B o

Ice- ackingDVL

30 m

90 m

- T an mi e an d ce , - Recei e h d o hone

GPS & f ee a eadio

2 km

Modem Range

• 3-4 ice moorings with 10 kHz µModemsabove and below Beaufort Lens

• UUV INS, upward-looking DVL

• CSAC/GPS synchronized modem TDMA

Integrated Navigation & Communication

• UUV Status Reports to topside

• 1-way travel times converted to range using embedded TDA

• Topside tracking solution• Autonomous selection of optimal

transmit modem

• Navigation drift and errors transmitted via Acomms to UUV together with

current ice motion• Onboard navigation fusion of INS,

modem range tracking and DVL,

constrained by UUV dynamic model.• Full system simulation using Virtual

Ocean Autonomy Testbed with HITL

39




Acoustic tracking system (LBL):

• UUV transmits a time-stamped status reports on TDMA schedule

indicating its current position & vehicle status.• µModems receive the message, & calculate OWTT from UUV to

each modem.

µModem Systems 1 & 2

µModem Systems 3 & 4

• OWTT & GPS coordinates of each modem, and last

reported UUV position is

used to triangulate the new position of the UUV.

• Uses Virtual Ocean for converting OWTT to range

estimate.

• Triangulation uncertainty is computed with a particle

filter.• New position delta, position

stdev and ice-camp drift

information is sent back to UUV.

µModems translate and with the ice sheet

40

2/23/21

20


Top-side (Ice camp)

AUV backseat (MIT's payload computer)

Micro modem

1

Topside Tracking

System

Acomms

Handler

Frontseat

interface

IXBlue PHINS INS

AUV frontseat

(manufacturer-side)

Teledyne RDi DVL

Navigation sensors

HydroMAN navigation system

ICEX Manager

Vehicle dynamic

model

Real-time model

calibration

Sensor fusion

DVL ice-drift

correction

LBL time lag

correction

Micro modem

2

Micro modem

3

Micro modem

4

IvP Helm

Vehicle-side acoustic

hardware

41


A sample LAMSS simulation run:



42

2/23/21

21


Modem Performance Metrics

1500 2000 2500 300020

40

60

80

100

Range (m)

Packet success (

%)

1500 2000 2500 300030

40

50

60

70

80

SN

R (

dB

)

rx: 30mrx: 90mrx: 30mrx: 90m

1500 2000 2500 300020

40

60

80

100

Range (m)P

acke

t su

cce

ss (

%)

1500 2000 2500 300030

40

50

60

70

80

MP

P (

dB

)

rx: 30mrx: 90mrx: 30mrx: 90m

HITL-NETSIM Virtual Ocean Missions ICEX20 Modem Range

Maximum SNR Maximum SNR

Multipath Penalty SNR Multipath Penalty SNR

43


ICEX203-hour Submerged Survey Mission

ITP 2013 ICEX2020

44

2/23/21

22


ICEX2020 Recovery

Tuesday, March 17, 2020

• ‘Steep’ Macrura/Durip recovery

45


LAMSS 1997 – 2017Lessons Learned

• Artificial Intelligence is critical to persistent and resilient operation of undersea distributed sensing systems

– Adaptation and collaboration may compensate for reduced sensor performance

– Communication channel inherently layered, highly band limited, latent and intermittent

– Integrated of sensing, modeling, and control required for sustained autonomous operation

• Nested, behavior-based autonomy is a key enabler – Nested modularity supports effective ’cloning’ of domain experts

– MOOS-IvP is open-source, highly portable autonomy software • Multi-objective optimization HelmIvP is key enabler for adaptive autonomy.• Provides 95%+ of leveraging autonomy software through nested repositories• Provides templates for efficient application and behavior development by domain experts

– Adaptive sensing, communication and autonomy supported by embedded environmental and tactical modeling

• Robust and Resilient Onboard Data Processing– Processing products suited for machine decision making

– False Alarm Control is critical. Ocean is random!

– Robustness more critical than resolution!

• Virtual Experiments key to deployment of robust and resilient field systems– Adaptive autonomy is inherently unpredictable. Robust and resilient performance requires extensive testing with

actual autonomy software

– Requires high-fidelity, physics-based environmental simulation (oceanography, acoustics, dynamics).

– 500-1000 times more hours spent in virtual experiments than real ones for distributed sensing concept development.

– Hardware-in-the-Loop support• Stimulation of embedded processing chain

• Analog modem transmit/receive support

46

mit marine robotics autonomous underwater vehicles

Documents