If your company is turning to continuous wave (CW) technology as an end-game LiDAR solution for your future passenger vehicles, robotaxis, or commercial fleet, there could be a not so tiny problem that many LiDAR manufacturers won’t tell you about.
Pointcloud blur, as we refer to it, is the blurring of objects in the pointcloud caused by rapidly moving mirrors designed to steer the laser across the environment. As the mirrors move, the rapid motion creates an effect that is very similar to an increase in beam size.
While blurring might seem like a small, insignificant issue to overcome, when you think about the bigger picture and the perception stack behind the pointcloud – and the fact that most manufacturers don’t have any potential solutions – even small amounts of pointcloud blur could lead to a sequence of issues that will ultimately prevent your vehicle from achieving the level of reliability and accuracy required to reach true Level 4 (L4) autonomous driving.
At Baraja, we’re proud to be the first to present a real solution to pointcloud blur and provide automakers with the LiDAR technology they need to power their L4 autonomous vehicles. Read on to learn more about what pointcloud blur is, how it happens and why it matters to you.
How does Pointcloud Blur Occur?
Legacy LiDAR designs that measure the time of flight with high peak power pulses have their limitations. Traditional MEMS based and spinner LiDAR sensors are usually bulky and the high peak power and high-resolution requirements often push the safety limits for the human eye. Now, the industry is looking to continuous wave technology for a high-performance, low power alternative that can provide a small form-factor, economical solution.
Many of these CW powered LiDAR sensors rely on polygon mirrors and MEMS based steering in the fast axis to steer the light over the field-of-view. However, due to the high rotational speeds and inertia, this means the scanner must constantly move over the duration of a single sample, causing the laser itself to be steered a measurable amount as well. So, instead of collecting an individual point in space, it unintentionally “blurs” the laser over a larger area, causing the pointcloud blur.
Much like if you were to take a picture of something while swinging your camera from one side to the other, the blurring of the laser creates a blurred, unclear image. You may be able to see that something is there if the reflectivity is high enough, but important details become distorted, making it difficult to identify what the object is, its dimensions and potentially losing smaller items altogether. In autonomous vehicles, this inaccuracy is unacceptable. LiDAR needs to serve as the “ground truth”, providing a complete and accurate depiction of its surrounding environment.
The problems of pointcloud blur
When it comes to LiDAR, having “blurred” data drastically impacts the accuracy and reliability of its output data. For some continuous wave LiDAR systems, in the fast axis of scanning, the scanner can move as much as 0.3° over a single measurement, significantly higher than the industry standard resolution requirement of 0.05°, which is needed for meaningful data collection at long ranges.
To put this in perspective, due to the rapidly moving fast axis scanner over 120°, and a typical sample duration of 4µs, the beam can be blurred as much as 0.52m at a 100m range, scaling to 1.04m at a 200m range.
This level of distortion would make it very difficult for automakers to properly identify what objects are in the pointcloud. At 200m, a pole could look the same as a person or a wall. This creates issues with the perception stack, where object detection and classification occur, ultimately impacting the vehicle’s ability to decide if it should stop, swerve, continue, or proceed with caution.
Solving the issues of pointcloud blur with true solid-state design
Pointcloud blur creates fundamental flaws in the autonomous architecture – and with many LiDAR manufacturers lacking a potential solution, it could mean they are investing in a design that simply won’t work.
At Baraja, we’ve developed the industry’s first true solid-state fast axis scanner design and blur-free LiDAR enabled by Spectrum-Scan™ technology. Instead of relying on a moving mirror to reflect light, we’ve harnessed the power of prism-like optics to turn one laser into the equivalent of hundreds, even thousands, of individual lasers.
Our unique patented LiDAR design has enabled us to maintain complete control over the fast axis scanning, enabling our laser to remain stationary in space and “hop” to each new sample versus “blurring” the laser from sample to sample.
So, instead of relying on oscillating mirrors, we simply refract light through a prism to scan, allowing our one laser to create up to 2000 individual direct points without any moving parts or “blurring” of light. With this, we are able to steer a low divergence laser beam in the fast axis at more than a million times per second with no-moving-parts.
This patented method not only eliminates the possibility of pointcloud blur, but it allows us to deliver exceptional resolution with less noise, fewer artefacts and fewer false positives – enabling L4 autonomous driving.
If you’re looking for a LiDAR with pointcloud precision that will take your vehicles to self-driving, invest in Baraja Spectrum-Scan™ technology. Our LiDAR system is years ahead of the competition and is solving the issues of tomorrow, today. Contact us now to learn more.