Radar Systems for Planetary Exploration

1
Radar Systems for Planetary Exploration

• Mike Taylor
• taylor_michael_a5_at_cat.com

2
Perception in Offroad Environments

• Offroad environment as well as robots themselves
  are very harsh on sensors and sensor performance.

3
Why Radar?

• Radar Positives
• Impervious to rain, mud, fog, dust.
• Few interference concerns
• Generally physically tough

• Radar Negatives
• Costly
• Wide beams
• Slow scan speeds
• Very hard to determine target size or shape
• False Alarms

• Key Points
• Radar provides a generally robust sensing
  solution.
• Sensor choice push against technology or push
  against physics.

4
Uses

• Object detection
• Terrain mapping
• Object tracking
• Sensor Fusion
• Camera radar
• Automotive groups exploring this area
• Laser radar
• Many robot systems use this. Boss, etc.

5
Basics

• Standard echo-location
• Radar emits specific radio frequency and detects
  reflected waves
• Separate transmit and receive antennas
• Single transmit/receive antenna
• Scan the beam to look in different directions
• Air traffic control radar
• Scanning determines the Field of View (FOV)
• Air traffic control radar 360
• Roving from North to East and back 90

6
RF Propagation

• Returned energy proportional to range-4
• Double the range, get only 1/16 the power
  returned
• RF propagation on transmit
• Same amount of power would hit each target
• Target 1 1 W m-2
• Target 2 ¼ W m-2
• Double the range, ¼ the incident energy
• P a range-2
• Reflected energy suffers same degradation
• Round trip range-2 range-2 range-4

Target 2 2 x Range 4 x Area
10 m
20 m
7
Beam Shape

• Beam shape is function of antenna
• High gain vs. omni-directional antennas
• Gain developed by interference
• Beam Width estimates
• 3 dB typical
• Contains vast majority of energy
• Relationships
• Beam Width a frequency -1
• Beam Width a (antenna width) -1
• Applies in both height and width

8
Side Lobes

• Result of same interference pattern that created
  the main beam.
• Generally much weaker than main beam.
• Objects receiving energy from side lobes can be
  detected.
• Car off to right as were driving down the road.
• Major issue for terrain mapping.
• Affect confidence of detection.
• See Alex Foessels PhD thesis for further
  discussion.

Image from appolo.lsc.vsc.edu
10 m
20 m
9
Beam Shape vs. Resolution

• Beam width affects angular accuracy and ability
  to separate targets
• Correlates to resolution
• Comparison to laser
• Laser beam size usually lt 1
• Radar beam size most 3 to 5
• Down sides to smaller beams
• Higher frequency vegetation opacity line of
  site
• Larger antenna hard to scan, larger form factor

Can the truck fit through?
Only one object reported with high angular error.
10
Radar Types
Frequency Scanning Data Output
24 GHz Mechanical Antenna or mirror motion Raw Data Powerful Resource intensive
77 GHz Phased / Patch Array - Issues with wide views Detections Range, Angle, Power Simple, Limited
94 GHz - Legal in U.S.? Return Processing Beam forming on return Dipole Tracks - Traffic use Limited

• Additional Specs
• Detection range for certain objects
• Horizontal and vertical beam width
• Horizontal and vertical FOV

• Scan rate
• Number of targets per scan
• Range and angle resolution

11
Continental ARS-300

• Long range, dual mode ACC-style radar
• Spinning cylindrical reflector
• Grating on cylinder causes different interference
  patterns
• Specs
• Long Range 200 m, 17
• Mid Range 60 m, 60
• Beamwidth 3 degrees
• Return limit varies by version
• Reference Information
• Tartan Racing publications
• Example of steered beam system
• Unique antenna design
• Emitted energy focused on a particular area
• Prone to ghost velocities
• Far reaches of FOV have limitations

12
Delphi ESR

• Long range, dual mode ACC-style radar
• Specs
• Long range 200 m
• Medium range 60 m
• Return limit varies by version
• ESR Electrically-steered radar
• Volvo S60 ESR Mobileye camera
• Launches in 2010
• Reference Information
• http//delphi.com/news/featureStories/fs_2008_06_0
  2_001/
• Example of beam forming on return
• Beam is not steered, wide emission pattern
• Bearing calculated by phase difference between
  multiple receive antennas
• Provides locations of returns above threshold
• Limits available information for processing

13
ACC Comments

• Cheap, useful, feature-filled radars
• Can be hard to acquire
• Limited to manufacturers tools and code
• Not tuned for offroad
• Incorrect thresholds
• Improper motion models
• Ghost Velocities
• Imperfect noise handling
• Wide beam angles
• Good first step

14
M/A-Com

• Low cost, low range radar for collision
  prevention and blind spot coverage
• Specs
• Single Mode
• Range 27 m
• FOV 100
• Limited returns
• Particularly good at picking up moving objects
• Reference Information
• http//www.macom.com/macom_prodnews.asp?ID1094
• Example of Dipole Radar
• Two receive antennas
• Returns signals are compared to determine bearing
• Potential ambiguity in bearing

Path length difference determines bearing
15
Angular Ambiguity

• Simple dipole radars have a weakness
• Both objects below are at roughly the same range
• Simple systems report seeing a single target
  along the centerline

16
NavTech

• Spin-off from ACFR
• Specs
• FMCW
• 360 Degree FOV
• 2 degree beam
• 2.5 Hz
• 0.03 meter range accuracy???
• 200 meter range
• Initial models could not measure velocity
• Reference Information
• http//www.nav-tech.com

17
FMCW Quirk

• Relative velocity causes vertical (frequency)
  shift in signal
• Range causes horizontal (temporal) shift in
  signal
• Up and down ramp allows separation of range and
  Dopper
• Up Delta R D
• Down Delta R - D

R D
R D
R - D
R - D
18
Other Suppliers

• Research Houses (for semi-custom radars)
• Militech
• http//www.millitech.com/
• Epsilon Lambda
• http//www.epsilonlambda.com/
• Manufacturers
• Eaton-Vorad
• http//www.roadranger.com/Roadranger/productssolut
  ions/collisionwarningsystems/index.htm
• Bosch
• http//rb-kwin.bosch.com/us/en/safety_comfort/driv
  ing_comfort/driverassistancesystems/
• adaptivecruisecontrolacc/index.html

19
Reflectivity and RCS

• All objects reflect energy. Two questions
• How much?
• In which direction?
• Units dBsm
• Reflected power relative to one square meter of
  flat metal sheet
• Human -10 to 0 dBsm
• Car 10 dBsm
• Energy reflected depends on
• Material
• Surface structure (clothing wrinkles)
• Size.
• Shape- Specularity

20
Radar Return vs. XY Position
21
Radar Tuning Scene
16 Rock
Senor Origin
6 Dia. Pipe
22
Radar Return vs. XY Position
23
Radar Return vs. XY Position
24
Radar Target Amplitude Curves
Key Trucks Human
Ground Noise
Amp(dB)

• Ground return is terrible
• Objects are specular
• Coke can challenge

Range(m)
25
Boss

• Vehicle Tracking
• Radar Lidar Fusion
• Direct velocity measurements key
• Orientation is challenging
• Veggie Cars

26
Motion Free Scanning Radar (consortium with CMU)
Motion Free Scanning Radar Sensor

• Narrow beam
• High reliability
• Low cost
• Small (30cm ? x 20cm L)

High Resolution Range Map
27
Cat AMT

• Radar-based autonomous mining truck (AMT) circa
  1995
• Millitech-developed 3D scanning FMCW radar
• Multi-sensor AMT under development with CMU

28
SSOD

• SSOD Slow Speed Obstacle Detection
• Blind spot detection system
• Option on some Caterpillar mining trucks
• M/A-Coms compliment WAVS in-cab camera system
• Turns off after short distance

29
(No Transcript)
30
Researchers ACFR

• Australian Center for Field Robotics.
• University of Sydney
• Rare radar research group
• Focused on mining applications
• Semi-stationary terrain mapping
• Assemble custom systems based on needs
• Purchase and fabricate components
• Develop own processing
• Paper repository
• http//www.cas.edu.au/publications

31
ACFR Radar Mapping

• Stope fill monitoring
• Filling large, mined out voids in underground
  mines
• Visibility very limited
• Fill monitoring as well
• Beam width 1.12
• 77 Ghz
• 30 cm range resolution

32
ACFR Radar Mapping

• Drag-line Monitoring
• Poor visibility limits productivity
• Provides situational awareness for operator
• Terrain
• Bucket
• Ropes
• Allows digging in zero vis

33
Researchers

• Steve Shedding, ACFR
• Former Postdoc at R.I
• Working in interesting mobile terrain mapping and
  map fusion
• Graham Brooker, ACFR
• Major push behind designing new radar systems at
  ACFR
• Alex Foessel
• R.I. PhD, now at John Deere company
• Research Houses
• Millitech
• Epsilon Lambda
• NavTech
• Automotive Suppliers

34
Improvements

• Lower Prices
• Automotive industry Delphi, Continental, Bosch
• Improved performance
• ACRF, automotive industry
• Sensor fusion
• Automotive, ACFR
• Delphi Volvo S60 ESR Mobileye
• Velodyne for radar
• ABM radar?

35
Radar Layout Method

• Calculate the number of radars required to cover
  all potential movement.
• Vehicle specs
• Top speed
• Minimum turning radius
• Minimum deceleration
• Calculate envelope
• Radar specs
• Field of view
• Detection Range
• Depends on target
• May vary with heading

36
Radar Layout Method

• Radar specs
• 60 m range
• 90 FOV
• This radar has sufficient range but insufficient
  FOV.
• Two radars will suffice

37
Homework

• Design a radar layout for a ground vehicle
  exploring a desert region
• Given
• Two radars
• Option 1 ACC-style, 5,000. 60 FOV, 150 m
  range for vehicles.
• Option 2 Raw data, 35,000. 90 FOV,
  90 m range for vehicles.
• Truck
• 12 meters long
• Rear differential is 2 meters from rear of
  machine
• 5 meters wide
• Turning radius 15 meter
• Top speed 12 m s-1 (assume
  independent of turning radius)
• Deceleration 1.5 m s-2
• Questions
• How many of each radar would you need to handle
  the vehicle?
• Which radar would you choose? Write a short
  blurb on why.
• Factors to consider number of sensors,
  adjustability, cost, computing and personnel
  resources.
• Assume your team is a typical CMU robotics team
  in the FRC with the normal skill sets, funding
  issues, and compressed timeline. There is no
  right answer- the key is going through the
  decision process and weighting each issue as you
  see fit.

38
References

• Textbooks Introduction to Radar Systems by
  Skolnik
• http//search.barnesandnoble.com/Introduction-to-R
  adar-Systems/Merrill-I-Skolnik/e/9780070579095/?it
  m4
• ACFR Publication Depot
• http//www.cas.edu.au/publications
• Overview of Delphi ACC systems including ESR
  Radar
• http//delphi.com/news/featureStories/fs_2008_06_0
  2_001/
• M/A-Com
• http//www.macom.com/macom_prodnews.asp?ID1094
• NavTech
• http//www.nav-tech.com
• Boss / Urban Challenge Papers (Continental
  radar)
• http//www.darpa.mil/GRANDCHALLENGE/TechPapers/Tar
  tan_Racing.pdf
• http//www.ri.cmu.edu/pub_files/pub4/darms_michael
  _2008_1/darms_michael_2008_1.pdf
• http//www.tartanracing.org/press/boss-glance.pdf