Stephen Crouch 19th Coherent Laser Radar Conference

CLRC 2018, June 18 – 21 1

Advantages of 3D Imaging Coherent Lidar for Autonomous

Driving Applications

Stephen Crouch Blackmore Sensors and Analytics, Inc.

136 Enterprise Blvd., Bozeman, MT, 59718

crouch@blackmoreinc.com

Abstract: Blackmore has developed the first real-time, 3D imaging lidar product based

on frequency modulation and coherent detection for autonomous driving applications.

Units began shipping to automotive customers in late 2017. The technology provides

long range performance (>300m), impressive point throughput (300k to 1.2M points

per second per scanned beam), total interference immunity, and a unique ability to

measure radial velocity on each 3D data point. In addition, continuous wave

transmission is an ideal fit for integrated photonics and solid-state scanning solutions

such as optical phased arrays. These advantages highlight an incredible market

opportunity for coherent lidar systems in automotive and other consumer applications

where direct detection lidar fails to meet system requirements. Technology building

blocks in high-throughput processing, integrated photonics and optical phased array

technology, all under development at Blackmore, will be presented in addition to

exciting mobile, geo-registered 3D point cloud data results.

Keywords: Coherent Laser Radar, 3D Imaging, Autonomous Driving, Range-Doppler Imaging

1. Introduction

Autonomous transportation systems have seen a spike in interest in recent years. Successful research

and pilot programs by industry leaders such as Waymo have demonstrated the possibility of autonomous

systems to make transportation safer, more efficient, and easier to access [1]. A variety of sensors

including cameras, radars, ultra-sonics, and lidar work together through complicated artificial

intelligence (AI) stacks to complete the necessary navigation and avoidance tasks. Of these sensors,

lidar is the most nascent and highest cost. Yet experts agree that lidar technology is critical to the success

of the “self-driving” revolution.

Automotive lidar market leader Velodyne distributes a lidar sensor based on pulsed, time-of-flight

ranging operated at 905nm wavelength. While this sensor has proven to be indispensable for this task,

the limited range performance (<100m typical), high cost, and issues with sunlight interference have

triggered a market for better lidar technology. According to released specifications, the pulse energy

per measurement of a Velodyne system is approximately 300nJ [2].

The most capable automotive lidar systems to-date have transitioned to 1550nm wavelength sources. At

this wavelength, the systems can emit significantly higher optical power [3]. This allows manufacturers

to make more measurements at longer ranges (~200m). While systems such as Luminar have succeeded

in hitting most automotive OEM specifications, the inefficient use of optical power via fiber laser

amplifiers underlines a questionable pathway to scalability. In some cases, manufacturers have claimed

to use in excess of 10uJ per pulse leading to time averaged optical powers of >10 watts per system [4].

In the search for optimal lidar solutions, many have dismissed coherent detection as “impossible" from

both technical and cost perspectives. However, Blackmore's products prove that coherent systems have

an important role in autonomous vehicle sensor suites. The presented technology represents a step for

coherent laser radar (lidar) into mass market applications.

2. Coherent Lidar Operational Advantages

Coherent lidar has several operational advantages that have been demonstrated in over 18 months of on-

vehicle, mobile lidar operation. These advantages include range performance, interference rejection, and

Mo9

Stephen Crouch 19th Coherent Laser Radar Conference

CLRC 2018, June 18 – 21 2

direct velocity measurement. The long-range performance of Blackmore’s lidar system is shown in

Figure 1 where the data was collected with a real-time processing and geo-registration.

Figure 1. A few seconds of lidar data is accumulated to demonstrate the fill factor and range of a

coherent lidar imaging system.

The range demonstrated in Figure 1corresponds to a system operating with a 100mW continuous wave

output power and 305kHz repetition frequency. The observable range extends beyond 450m for brighter

diffuse targets. System parameters are included in a proprietary link budget model for a 10% diffuse

reflector which shows good agreement with traditional models and supports good SNR at a 200m range

[5-7].

Figure 2. Blackmore’s link budget model is compared to classic coherent detection papers and shows

good agreement.

Coherent systems are very sensitive to the doppler effect. While operating with this effect can be

technically challenging, signal processing solutions enable excellent velocity precision (<0.25 m/s) over

very short measurement time intervals (3.3 us). The rapid measurement of range and velocity sets this

approach apart from automotive radar systems that need to operate on timescales of tens of millisecond

to effectively resolve similar Doppler frequencies. Example Doppler imagery is shown in Figure 3.

Vehicle Location

In-N-Out Burger Sign

Stephen Crouch 19th Coherent Laser Radar Conference

CLRC 2018, June 18 – 21 3

Figure 3. (TOP) A series of frames is from Blackmore’s automotive lidar system is shown. Points are

colored by velocity showing clear resolution of motion components on the moving target. (BOTTOM)

Mobile lidar data showing pedestrian tracks when colored by the Doppler data field.

The Doppler field allows more reliable measurement of scene activity with lower latency. This

translates into more efficient artificial intelligence for vehicle navigation by using velocity-based scene

segmentation, tracking, and object identification. Ultimately this increases the measurement certainty

and overall “value” of each 3D data point.

3. The Pathway to Scalability

The discussion in sections 1 and 2 highlights a stark contrast between the optical power requirements

for pulsed/direct-detect and CW/coherent scanned lidar approached. The >10x difference in optical

power efficiency of coherent detection on a per measurement basis is an important consideration in

discussing the scalability of lidar designs. High optical powers will not translate to small, mass

manufacturable chip designs.

Optical fiber communications points to a solution. This space has long targeted coherent detection as

the preferred implementation for high bandwidth, long-haul, power efficient systems [8]. Many of the

same technology building blocks such as narrow linewidth lasers and balanced receivers, along with

broader supply chain capacity, can be leveraged in the construction of scalable coherent lidar systems.

An example of InP chips from on of Blackmore’s recent wafer runs is shown in Figure 4. These chips

represent a major subsystem for future automotive lidar design at Blackmore and a step towards “lidar

on a chip”.

Non-mechanical beamsteering is often cited as a long-term requirement for automotive lidar. The low

optical power requirements offer a match with integrated photonics for non-mechanical beamsteering

where high peak powers are not acceptable. Blackmore is working with Sandia National Labs to develop

silicon photonics based non-mechanical beamsteering devices for use in future systems. While these

systems are indeed challenging to realize and require special attention to insertion loss mechanisms, the

opportunity to realize complete “lidar chipsets” is clear. Ultimately, lidar sensors need to be chip-scale

to ever realize the low cost and volume potential of the automotive market.

Stationary Receding

+3.5m/s -3.5m/s

Illumination Direction

Velocity = Doppler

Stephen Crouch 19th Coherent Laser Radar Conference

CLRC 2018, June 18 – 21 4

Figure 4. A set of InP chips with several lidar subsystem building blocks is shown in testing. The

chips were developed in conjunction with Fraunhofer HHI and JePPIX.

As a final technical point, the author notes that coherent detection has significantly less susceptibility to

interference from sunlight or other lidar system due to spatial, spectral, temporal, and other filter options

that make interference an impossibility. As automotive lidar scales to the broader market, this advantage

will prove to be invaluable as interference rejection in coherent lidar is inherent and does not require

expensive optical bandpass filters which may degrade across temperature and field of view. In a safety

critical application, performance must be guaranteed across the broadest possible operational set. This

includes immunity to interference.

4. Conclusions

The requirements for automotive lidar are challenging for any optical technology to address. However,

coherent detection offers a unique set of capability and scalability advantages that will support

significant future investment as other technologies fail. The low optical power requirements relative to

pulsed, direct-detect systems will be the major differentiator as the automotive industry pivots to

coherent lidar. Through prototype systems and technology investments, Blackmore has demonstrated a

roadmap for driving coherent laser radar towards mainstream applications.

5. References

[1] T. Higgins and C. Dawson, “Waymo Orders Up to 20,000 Jaguar SUVs for Driverless Fleet,” Wall Street

Journal, 27-Mar-2018.

[2] Velodyne. “HDL-64E, Resource Manual, Laser Safety Parameters,” November 7, 2007.

[3] “Encyclopedia of Laser Physics and Technology - eye-safe lasers, retina, corneal injuries, erbium, thulium.”

[Online]. Available: https://www.rp-photonics.com/eye_safe_lasers.html.

[4] Jeff Hecht. “Under the Hood of Luminar's Long-Reach Lidar.” IEEE Spectrum. July 5, 2017.

[5] R. G. Frehlich and M. J. Kavaya, “Coherent laser radar performance for general atmospheric refractive

turbulence,” Appl. Opt., AO, vol. 30, no. 36, pp. 5325–5352, Dec. 1991.

[6] C. M. Sonnenschein and F. A. Horrigan, “Signal-to-Noise Relationships for Coaxial Systems that Heterodyne

Backscatter from the Atmosphere,” Appl. Opt., AO, vol. 10, no. 7, pp. 1600–1604, Jul. 1971.

[7] J. Y. Wang, “Detection efficiency of coherent optical radar,” Appl. Opt., AO, vol. 23, no. 19, pp. 3421–3427,

Oct. 1984.

[8] F. Corporation, “Finisar Introduces Industry’s Smallest Coherent Optical Assembly for High-Density Line

Card and Transceiver Designs at OFC 2018,” 13-Mar-2018. [Online]. Available: http://globenewswire.com/news-release/2018/03/13/1421255/0/en/Finisar-Introduces-Industry-s-Smallest-Coherent-Optical-Assembly-for-High-

Density-Line-Card-and-Transceiver-Designs-at-OFC-2018.html.