Localization and Mapping in Confined Areas with a Hovering AUV - - PowerPoint PPT Presentation

▶
localization and mapping in confined areas with a
SMART_READER_LITE
LIVE PREVIEW

Localization and Mapping in Confined Areas with a Hovering AUV - - PowerPoint PPT Presentation

Dec 11, 2022 19 likes •456 views

Marine Robot Localization and Navigation Workshop @ ICRA 2016 Localization and Mapping in Confined Areas with a Hovering AUV Michael Kaess Robotics Institute Carnegie Mellon University May 20, 2016 Inspecting Ships and Harbor Infrastructure

slide-1
SLIDE 1

Marine Robot Localization and Navigation Workshop @ ICRA 2016

Localization and Mapping in Confined Areas with a Hovering AUV

Michael Kaess Robotics Institute Carnegie Mellon University May 20, 2016

slide-2
SLIDE 2

Michael Kaess 2

Inspecting Ships and Harbor Infrastructure

SS Curtiss, San Diego Drift-free navigation + ensure full coverage Slow for covering large areas For non-complex areas we use imaging sonar instead

slide-3
SLIDE 3

Michael Kaess 3

Why Sonar?

Camera Sonar

slide-4
SLIDE 4

Michael Kaess 4

Collaborators

  • MIT

– John Leonard, Franz Hover, Pedro Teixeira, Josh Leighton

  • Univ. Michigan

– Ryan Eustice, Matt Johnson-Roberson, Paul Ozog, Stephen Chaves, Jie Li

  • CMU

– Tiffany Huang, Eric Westman

slide-5
SLIDE 5

Michael Kaess 5

HAUV: Hovering Autonomous Underwater Vehicle

Equipped with:

– 5 Thrusters – Battery (1.5 kWh) – Ring laser gyro – Sonars:

  • Doppler Velocity Log (DVL)
  • Multi-beam sonar
  • Both are actuated

– Cameras:

  • Stereo with LED lights
  • Periscope

– Fiber tether

HULS3

slide-6
SLIDE 6

Michael Kaess 6

Recent Experiments (Confined Area Search)

Mar+Jun 15: SS Curtiss, San Diego (180m) Aug 13: USS Saratoga, Newport (324m) Aug 14: NS Savannah, Baltimore (180m)

slide-7
SLIDE 7

Michael Kaess 7

Curtain Mission for Complex Areas

slide-8
SLIDE 8

Michael Kaess 8

Close-up Inspection to Resolve Small Structures

  • 1. De-noising sonar data
  • 2. Eliminate drift using feature-based navigation (FBN)
slide-9
SLIDE 9

Michael Kaess 9

De-noising

  • Difficult to extract range

information because of artifacts

  • Several sources of error make

this a difficult task

– Cross-talk / side lobes – Reflections/multipath – Vehicle motion – Noise (ambient & electrical) – Target structure

slide-10
SLIDE 10

Michael Kaess 10

DIDSON: Operating Principle

  • DIDSON operating mode:

– 8 cycles to build a scan – 12 transducers fire simultaneously in each cycle – Interleave results to obtain the complete scan – 10 frames/s → 12ms/cycle

  • Although the motivation for

this operating mode is to reduce cross talk by not firing adjacent transducers simultaneously, there is still significant cross talk:

  • Transducers have finite (non-

zero) gain at the main lobes of

  • ther transducers in the same

cycle!

slide-11
SLIDE 11

Michael Kaess 11

Assembling a Point Spread Function

  • A 1-dimensional PSF

can capture this angular dependence

  • Need beam pattern

for transducers

  • Assuming invariance

we can use a single beam’s beam pattern (e.g.) center beam

slide-12
SLIDE 12

Michael Kaess 12

Experimental Validation

Angular Radial (ignored) Sonar image (0.8mm target)

slide-13
SLIDE 13

Michael Kaess 13

slide-14
SLIDE 14

Michael Kaess 14

Filtering – Improved Resolution

  • Small object becomes visible
slide-15
SLIDE 15

Michael Kaess 15

De-noising – Improved Resolution

slide-16
SLIDE 16

Michael Kaess 16

FBN with Planar Surfaces

  • The map consists of

(infinite) planar surfaces

  • Pool experiment:

Kaess, ICRA 2015

Real-time with a handheld RGB-D (Kinect) sensor

slide-17
SLIDE 17

Michael Kaess 17

Beyond Planes: FBN with Submaps

  • Sequential pings do not overlap
  • Accumulate submaps (low drift over tens of seconds)
  • Alignment produces pairwise pose-to-pose constraints
  • Integrated with vehicle navigation in factor graph
  • Online solution by iSAM [Kaess et al., IJRR 12]
slide-18
SLIDE 18

Michael Kaess 18

Pool Experiment

Dead reckoning FBN

slide-19
SLIDE 19

Michael Kaess 19

  • FBN eliminates long-term drift

FBN with Submaps

slide-20
SLIDE 20

Michael Kaess 20

Pier (PAX River 2015)

Dead reckoning FBN

slide-21
SLIDE 21

Michael Kaess 21

Non-Complex Area: Imaging Sonar

Hull relative navigation, 1.5m standoff distance

slide-22
SLIDE 22

Michael Kaess 22

Imaging Sonar Registration

Frame A Frame B

slide-23
SLIDE 23

Michael Kaess 23

Imaging Sonar and FBN

  • State-of-the-art requires planar

assumption

  • Can we recover 3D geometry

from forward-looking sonar images?

  • Also want to recover vehicle

motion (feature-based navigation, FBN)

slide-24
SLIDE 24

Michael Kaess 24

Sonar Geometry – Unknown Elevation

Measured: Range r and bearing ψ Unknown: Elevation θ within opening angle of sonar,

e.g. 28° for DIDSON

rmin rmax

slide-25
SLIDE 25

Michael Kaess 25

Camera Geometry – Unknown Range

Measured: Image coordinates (u,v) related to bearing ψ

and elevation θ

Unknown: Range r

slide-26
SLIDE 26

Michael Kaess 26

  • Can recover 3D geometry from multiple views
  • Correspondence problem + Geometry recovery

Multiple Views: Structure from Motion

slide-27
SLIDE 27

Michael Kaess 27

Acoustic Structure from Motion (ASFM)

Elevation of a feature can be recovered from multiple views!

Ping 1 Ping 2 Ping i

slide-28
SLIDE 28

Michael Kaess 28

Factor Graph Representation

Bipartite graph with variable nodes and factor nodes

Robot pose Landmark position Landmark measurement Odometry measurement

slide-29
SLIDE 29

Michael Kaess 29

Nonlinear Least-Squares

Repeatedly solve linearized system

Efficient solution in online setting possible with iSAM (Kaess et al. 2008) and iSAM2 (Kaess et al. 2012)

argmaxΘ 𝑞𝑗(Θ)

𝑗

A

𝜖ℎ𝑗 𝜖Θ

Θ

  • b

ℎ𝑗 (Θ ) argminΘ ℎ𝑗 Θ

Ξ 2 𝑗

Gaussian noise argmin𝜄 𝐵𝐵 − 𝑐 2

poses landmarks

slide-30
SLIDE 30

Michael Kaess 30

  • Structure recovered with low uncertainty

Simulation – General Motion

Huang and Kaess, IROS 2015 Side view!

  • Three views
  • Known data association
  • 100 Monte Carlo runs
slide-31
SLIDE 31

Michael Kaess 31

  • Ambiguity: Cannot distinguish positive/negative

elevation angle

Simulation – Forward Motion

Huang and Kaess, IROS 2015

slide-32
SLIDE 32

Michael Kaess 32

  • Ambiguity: High uncertainty along arc

Simulation – Yaw and Sideway Motion

Huang and Kaess, IROS 2015

slide-33
SLIDE 33

Michael Kaess 33

  • Lower uncertainty: Roll disambiguates sign of elevation

Simulation – Roll Motion

Huang and Kaess, IROS 2015

slide-34
SLIDE 34

Michael Kaess 34

  • Forward/pitch motion provides best constraints,

followed by roll

Simulation – Summary

Huang and Kaess, IROS 2015

slide-35
SLIDE 35

Michael Kaess 35

Data Association is Difficult!

Points are ordered by range!

slide-36
SLIDE 36

Michael Kaess 36

Moving sonar changes order of projections

Data Association is Difficult! (2)

slide-37
SLIDE 37

Michael Kaess 37

Epipolar Geometry?

Projecting samples over elevation range to find putative correspondences

slide-38
SLIDE 38

Michael Kaess 38

Imaging Sonar – Boston Harbor

slide-39
SLIDE 39

Michael Kaess 39

Imaging Sonar - Manually Selected Features

slide-40
SLIDE 40

Michael Kaess 40

Imaging Sonar – Reprojection Error

Feature points Projected estimated structure

slide-41
SLIDE 41

Michael Kaess 41

Imaging Sonar – 3D Geometry

  • Before optimization - elevation of all features approx. equal
  • After optimization - elevation takes ladder structure
slide-42
SLIDE 42

Michael Kaess 42

Imaging Sonar: Ladder

  • Front view before (left) and after (right) optimization:

Elevation is clearly recovered (with some uncertainty)

slide-43
SLIDE 43

Michael Kaess 43

Future Work

  • Further improvement of model fidelity from profiling
  • Point feature extraction for ASFM
  • More dense structure recovery
  • Multiple AUVs