IEEE Spectrum October, 2016 - 14

LIdar ThaT wILL Make
seLf-drIvIng Cars
affordabLe
Cutting the cost of this sensor is the key to
a driverless-car future
Chances are you've never seen a
fully autonomous self-driving car
out on the street. But if you have, you probably couldn't help but notice the distinctive spinning sensor adorning its roof.
It's what helps an autonomous vehicle
understand the world around it, but it's
also what's keeping autonomous vehicles
from being affordable enough for the average consumer to buy. Lidar (light radar)
is complex and expensive right now, but
within the next few years, it'll be cheaper
and more reliable-and everywhere.
Lidar is a sensing technology similar
to radar that detects objects with pulses
of laser light. Though lidar has a shorter
range than radar (tens to hundreds of
meters), the much shorter wavelengths
used by lidar result in a massive increase
in resolution. The uniquely reliable and
high-quality data that lidar provides
has made it the sensor of choice for the
majority of vehicle autonomy applications. In fact, many experts feel that lidar
is a necessity for driverless cars.
"Lidar is a critical sensor for driverless vehicles because it enables highly
precise and robust localization across
a wide range of conditions," explains
Karl Iagnemma, CEO of nuTonomy,
a Cambridge, Mass., startup that is
currently testing self-driving cars in
Singapore. But Iagnemma points out that
the size, complexity, and cost of the current generation of lidar sensors are significant obstacles to the commercialization
of any technology that depends on them.
Many autonomous cars have relied on
the HDL‑64E lidar sensor from Silicon
Valley-based Velodyne, which scans
14

OCT 2016

NORTh AMERICAN

2.2 million data points in its field of view
each second and can pinpoint the location of objects up to 120 meters away
with centimeter accuracy. But the sensor
itself weighs more than 13 kilograms and
costs US $80,000. This year, Velodyne
announced the VLP‑32A, which offers a
200-meter range in a 600-gram package.
With a target cost of $500 (at automotive
scale production), the VLP‑32A would

PixelS And PointS: A Velodyne lidar
produces a set of data points that show nearby
cars. Makers of self-driving cars consider it
essential tech.

be two orders of magnitude cheaper
than its predecessor but still too expensive to be integrated into driverless cars
intended for the consumer market.
A substantial amount of recent academic
and industry research has been focused
on making lidar sensors smaller, easier
to manufacture, and cheaper. At the CES
2016 electronics show, Quanergy Systems,
in Sunnyvale, Calif., demonstrated a prototype solid-state lidar sensor designed for
driverless cars. It uses an optical phased
array to steer laser pulses rather than a
rotating system of mirrors and lenses.
Quanergy projects that its sensor will cost
$250 in volume production, and it should

SPECTRUM.IEEE.ORG

be available to automotive original equipment manufacturers in early 2017.
Meanwhile, two startups are working on $100 automotive lidar systems
they both say will be released in 2018.
Innoviz, in Israel, is promising a "highdefinition solid-state lidar" with better
resolution and a larger field of view than
those in existing sensors. Innoluce, in
the Netherlands, is using a microelectromechanical mirror system to scan and
steer a laser beam instead of the solidstate approach; its engineers claim it
will outperform optical phased arrays
in both range and resolution.
One of the most exciting recent developments in lidar technology comes from
MIT. With funding from the U.S. Defense
Advanced Research Projects Agency's
Electronic-Photonic Heterogeneous Integration program, MIT researchers have
leveraged silicon photonics to condense
a functional lidar system onto a single
0.5- by 6-millimeter chip that can be fabricated in commercial CMOS foundries.
MIT's prototype has a range of just a few
meters, but there is a clear development
path toward a 100-meter range and a perchip cost of just $10.
The market for a small and cost-effective
solid-state lidar system is immense. In
August, Quanergy closed a $90 million
funding round on a $1.59 billion valuation,
while Ford and Chinese search giant Baidu
jointly invested $150 million in Velodyne.
The goal of these investments is to make
a $100 automotive lidar sensor available
within the next few years.
While the automotive industry represents the largest near-term market for
low-cost lidar, the availability of small
and inexpensive lidar sensors will benefit many other industries. Lidar in mobile
phones will enable high-resolution gesture recognition. Robots with lidar in
their fingers will recognize and manipulate objects with more confidence than
ever before. And lidar will give drones
the data they need to avoid obstacles-
necessary for safely performing tasks
like package delivery.
-eva n ack er m a n

Velodyne

NEwS

http://SPECTRUM.IEEE.ORG

Table of Contents for the Digital Edition of IEEE Spectrum October, 2016