Blackmore Doppler LIDAR: Is this what's missing from autonomy?
LIDAR is generally considered the base ingredient of autonomous driving: although Elon Musk may disagree, the vast majority of automakers believe that the only way a car can be driven autonomously, yet reliably, is by employing LIDAR sensors. Now, a company with years of experience and funding from BMW i-Ventures and Toyota AI Ventures claims to have reinvented the LIDAR - but what exactly has Blackmore done?
What does LIDAR do?
Traditionally, the laser and radar combination allows a 3D imaging of a vehicle's surroundings. This is made possible through the bouncing of laser beams off of nearby objects - as laser is used, LIDAR allows for a far-reaching field of view. Most importantly, it remains unaffected by changes in ambient lighting. Accordingly, dark conditions, fog, and rain, which make camera-based systems less effective, are no problem for LIDAR. The highly detailed, 3D map created using LIDAR allows vehicles to confidently navigate their way round. The data collected through camera-based systems does not offer the same level of detail; the lacking image often leads to peculiar accidents involving plainly painted trucks, fire-trucks, and other, stationary items.
Despite their imaging superiority, LIDAR systems are not without flaws. Firstly, the cost of LIDAR systems is prohibitive in some cases: for example, Velodyne, the go-to provider of LIDAR systems, sells its top-of-the-range system (VLS-128) for a reported $12,000 at volume. Secondly, LIDAR systems tend to be quite bulky: although some systems now use software to steer laser beams (solid-state LIDAR), the majority of them use hardware to physically spin the device, contributing to their larger weight and more complex manufacturing process.
How does Blackmore's Doppler LIDAR differ?
More importantly, LIDAR is -was- not able to calculate a moving object's velocity, at least not directly. In order to assess an object's speed, two consecutive laser beams had to hit the same item: the returning laser beams communicated the object's position at two different time points, which then, through calculation, allowed the vehicle to estimate the speed and direction of the object. As laborious as this may sound, it happened quite instantaneously. Not effortlessly, though: for all their accuracy, laser beams do not always manage to reach the same item twice, particularly when it comes to items in the distance; moreover, the data received needed to be processed in order to provide the final piece of information actually needed (read: speed), which contributed additional latency.
Blackmore now has a solution to this "inconvenience", taking advantage of its Frequency Modulated Continuous Wave LIDAR. By taking advantage of a phenomenon known as the Doppler effect, along with the "coherent" nature of its FMCW device, it manages to use a single shot of laser to secure all the information needed, including velocity and location of the object. Although the details of how the Doppler effect exhibits itself in wave physics is outside the scope of this article, it suffices to say that you can observe it when an ambulance (or any other vehicle sporting a siren) is moving towards or away from you: have you noticed how the sound of the siren changes as the ambulance drives away? The change in the pitch of the siren does not actually happen, yet we perceive it as happening. The same logic underpins Blackmore's technology: the change in the wavelength of the returning laser beam contains information about the velocity of the object the laser beam interacted with during its trip.
So, is Doppler, dope-r?
Although the Doppler LIDAR seems almighty, it still suffers from a few shortcomings: the most important one, as Alex Davies of Wired points out, is its inability to detect items moving exactly perpendicular to the projected beam. Such a "deficiency" may sound more serious than it actually is, as the density and multi-directional nature of the laser beams projected by the LIDAR will probably eliminate any chances of an item going by unnoticed. With that said, full autonomy needs to guarantee effectively flawless driving, so I would expect to see more data on the percentage of objects going undetected before weighing the pros and cons of this new type of LIDAR. For all that counts, the market definitely seems invested in Blackmore's technology: through a Series-B financing round led by BMW i-Venture, Blackmore recently raised $18m.
The future of LIDAR and autonomy seems more exciting than ever.
Photo Credit: Blackmore Inc