Developing the Aurora Driver safely for trucks requires that our perception system detect and track objects farther, faster, and with greater precision than ever before. Last year, we acquired Frequency Modulated Continuous Wave (FMCW) lidar, the game-changing technology that makes this possible.
For context, self-driving systems have traditionally been outfitted with AM lidar. AM lidar works by emitting brief light pulses at a fixed frequency. We can determine the location of objects based on how long it takes for those laser pulses to bounce off surfaces and return to the sensor. The farther away an object is, the longer it takes for the light to return.
In contrast, FMCW lidar sends out a constant stream of light (“continuous-wave”) and changes the frequency of that light at regular intervals (“frequency-modulated”). This allows us to both determine the location of objects and precisely measure their velocity using the Doppler effect.
FMCW lidar provides substantial advantages over even the most advanced AM lidar systems available today, especially when it comes to long-haul trucking. Read on to learn how FMCW lidar will help us deliver the Aurora Driver safely, quickly, and broadly.
Our FMCW lidar sensors already give us significantly better range performance when compared to typical AM lidar systems, allowing the Aurora Driver to see well beyond 300 meters even on targets that don’t reflect much light (think joggers in dark clothing). FMCW lidar is so powerful because the sensors are single photon sensitive, meaning they can detect the smallest amount of light possible. Aurora’s FMCW lidar also operates around the 1550nm wavelength band, allowing the sensors to broadcast stronger light pulses while still meeting eye safety standards. In contrast, other lidar systems are often not single photon sensitive and operate near the 900nm wavelength band, requiring them to limit their light output for eye safety reasons.
When the Aurora Driver sees farther, it has more time to react to unexpected obstacles. Even fractions of a second can make a huge difference in safety and comfort, especially with heavy vehicles driving at highway speeds.
Our most recent upgrades to Aurora Lidar allow our perception system to see objects that are more than 300m away. Grid lines in the video are 25 meters apart.
Unlike AM lidar, FMCW lidar provides very accurate velocity for each data point instantaneously. This helps our perception system process incoming data faster because it no longer has to estimate velocity from changes in object position.
Instantaneous velocity also makes it easier for the perception system to recognize distant and sparse data points as objects and track how those objects are moving over time. For example, an FMCW lidar sensor may only receive a few returns on an object 300m away, but if those hits give a velocity value of interest (e.g., moving towards the vehicle at >70 mph) our perception system knows the object is important because it is approaching quickly.
Faster identification and tracking gives the Aurora Driver more time to maneuver. A better understanding of how fast objects are moving also allows the Aurora Driver’s motion planning system to plan a better reaction.
Points are colored by Doppler. White points are static or stationary, blue and green points are moving toward the sensor, and red and yellow points are moving away from the sensor. As cars pass by the sensor, they transition from moving towards (blue/green) to away (yellow/red).
Traditional AM lidar systems can be affected by:
All of these events can cause errors in the data, resulting in problems like “ghost” objects that don’t actually exist or objects that are reported in the wrong location. Vehicles using AM lidar need extra hardware, complex software, and more computational power than FMCW lidar to manage these types of data artifacts.
FMCW doesn’t suffer from these types of issues because each sensor is specially designed to respond only to its own light pulses. If the returning light does not match the timing, frequency, and wavelength of what was originally transmitted, the FMCW sensor knows to filter out that data point. That means FMCW lidar enables more accurate object detection while also saving the hardware, software, and computational power traditionally used to mitigate AM shortcomings.
FMCW lidar gives the perception system more accurate data with less effort than AM lidar, enabling safer and smoother driving.
As more self-driving vehicles show up on the road, those powered by the Aurora Driver will not have to contend with interference issues from sensor crosstalk. FMCW lidar can also leverage technological and manufacturing advancements from other photonics industries, where components are continually evolving to be more robust and low-cost. Finally, because FMCW lidar sensors use less optical peak power than conventional AM pulsed lidar sensors, many of the photonics components for an FMCW lidar can be produced on a single chip. These chips can be manufactured cost-effectively in very large quantities and can be easily calibrated using automated software.
The Aurora Driver’s sensor suite will be scalable and affordable while providing cutting-edge performance.
FMCW lidar will accelerate the Aurora Driver’s ability to safely move both people and goods. It will also play a key role in manufacturing and deploying the Aurora Driver at scale. We’re integrating Aurora FMCW lidar into our test fleet this year, and we’ve begun training our perception system to work with FMCW data. We’re already seeing impressive results, and this is just the beginning–we expect to see even better performance as we continue to refine the system.
Want to help us develop groundbreaking technologies for the Aurora Driver? Check out our open positions and learn more about what it’s like to work at Aurora.
With half a dozen locations and counting, the Aurora Test Site Network allows us to efficiently test, validate, and make rapid progress on the Aurora Driver in a variety of geographies near our growing areas of development.
We’re excited to debut Aurora’s FirstLight Lidar as a sensor on Aurora’s next-generation test vehicles, where it helps our perception system see and track objects farther, faster, and with greater precision than ever before.
At Aurora, we’ve adopted a “smarter, not farther” approach to on-road testing.