For the full video of this presentation, please visit:
https://www.embedded-vision.com/platinum-members/nxp/embedded-vision-training/videos/pages/may-2018-embedded-vision-summit-roy
For more information about embedded vision, please visit:
http://www.embedded-vision.com
Arunesh Roy, Radar Algorithms Architect at NXP Semiconductors, presents the "Understanding Automotive Radar: Present and Future," tutorial at the May 2018 Embedded Vision Summit.
Thanks to its proven, all-weather range detection capability, radar is increasingly used for driver assistance functions such as automatic emergency braking and adaptive cruise control. Radar is considered a crucial sensing technology for autonomous vehicles not only for its range finding ability, but also because it can be used to determine target velocity and target angle. In this tutorial, Roy introduces the basic principles of operation of a radar system, highlighting its main parameters and comparing radar with computer vision and other types of sensors typically found in ADAS and autonomous vehicles.
After examining the features and the limitations of current automotive radar systems, Roy discusses how automotive radar is evolving, particularly in light of safety performance assessment programs such as the European New Car Assessment Programme (eNCAP). He concludes with a discussion of how radar systems may compete with or complement vision-based sensors in future ADAS-equipped and autonomous vehicles.
6. • Measurement Concept – FMCW (Frequency Modulated Continuous Wave)
• Carrier Frequency
• 24 GHz
• 77 GHz
• RF Power
• Output power limit regulated (ex: 10mW in Japan)
• Automotive radar range (<60m: SRR, 60-150m MRR, >150m LRR)
Model Year
Volume
77 GHz24 GHz
TREND
7. • Antennas – Patch antennas on PCB (printed circuit board)
• Patch antennas have enabled more cost-effective solutions
• Design provides a directional beam with side lobes (not shown below)
• Electronic Components
• Two IC packages or one IC package for primary functionality
• Plus support components (power supply, communications, EMC, etc.)
• System complexity reduction = costs savings (Gen1 ~ $1K, Gen4 ~ <$50)
9. • Radar is an active sensor providing 4-
dimensional attributes
• Range
• Maximum detection range, range resolution
• Velocity/Doppler
• Maximum detection velocity, velocity
resolution
• Angular/Azimuth
• Angle Resolution
• Elevation
• Single dimension optimizations can lead
to trade-offs in others
11. •
• Multiple Rx channels can be used to improve angular
resolution
•
•
•
•
•
•
12. •
• Range & Doppler measurement
inherent in technology
• System costs reducing leading to an
increase in the attach rate
•
•
•
•
•
Module
Size
Range Cost
RADAR
Vision
13.
14. • Quest for higher resolution
• Highway autopilot limitations today
• Driving higher requirements tomorrow
• Obstacle detection (>300m)
• Obstacle size (soda can)
• Higher requirement for angular resolution
• Optimizing for maximum detection range &
higher range resolution
• System optimizations in one dimension
can lead to trade-offs in others.
• Multiple sensor types may be required
15. Receiver Chain
Down Conversion
RADAR MCU
- Control
- Signal Processing
- Object Detection
- Object Classification
Signal Generation
Transmitter Chain
Receiver Chain
Down Conversion
Receiver Chain
Down Conversion
Receiver Chain
Down Conversion
• Multiple Rx channels can be used to improve angular resolution:
• Path length difference leads to phase shift of received signals
• Measured phase shifts used to calculate the “angle of arrival”
• Comparison to other sensor types (cost competitive with <2º resolution)
16. Rx Signal analysis Detection
Signal
conditioning
Tracking
High Performance MCU
•Antenna
•Mixer
•LNA
•HP/LPF
•AAF
•ADC
MMIC
Obj1=(d1, vr1,al1)
Obj2=(d2, vr2, al2)
Tracking
(e.g., Kalman + Munkres)
Range
FFT
Doppler
FFT
MIMO &
Beamforming
Adv Functions
Thresholding (ex
CFAR), 3D Interpolation
DOA
Super Resolution
(e.g., MUSIC,
ESPRIT)
Detection 3 Targets
• ~0.5m
• ~7.0m
• ~10.5m
•
24. Radar Vision
As resolution with radar improves, we enable a redundant, safer vehicle architecture
High Resolution RADAR Vision
Mapping
Classification
25. • Autonomous vehicle will not have a
single sensor type
• Radar provides key attributes to the
future of the autonomous vehicle
• Vison & radar are complimentary
• Further industry improvements will
enable:
• Higher performance systems
• Redundancy leading to safer systems
• High resolution radar & vision sensors
will continue to play a vital role