When it comes to fully autonomous driving, Doppler-enabled long-range LiDAR is a must. This combination leads to LiDAR sensors that are not only capable of long-range detection and measuring velocity – which is not much different from what most radars on the market are already accomplishing – but can also provide enough resolution to separate real returns from the noise and properly detect and identify individual objects within the field-of-view (FoV). However, beware that not all Doppler-enabled LiDAR sensors are ready to hit the road just yet.
Most LiDAR systems today are incapable of measuring Doppler. Instead, they rely on computationally-intensive frame-to-frame estimations of velocity – and those that are actually Doppler-enabled technologies, while gaining the ability to measure velocity per point, are in their infancy and need to trade off performance in other critical areas required for automotive LiDAR applications in order to achieve it.
At Baraja, we’ve paired per-point Doppler with our patented Spectrum-Scan™ technology to develop a winning solution LiDAR for automakers who want to enable true Level 4 autonomous driving capabilities with data they can trust. The Baraja LiDAR is capable of measuring the velocity of every point in the FoV without sacrificing resolution and reliability – so that your perception algorithms can respond quickly and confidently in any driving scenario.
When light bounces off of a moving object, say a moving car, a slight shift in the frequency of that light occurs. If the object is moving towards you, the frequency will increase. If it is moving away, it will decrease. This is known as the Doppler effect. This is the same phenomenon that happens when you hear the sound of a police car in pursuit. As the car moves towards you, the pitch of the siren appears to become higher, but as it passes by and moves away from you, the pitch appears to become suddenly lower.
Measuring this frequency shift, or Doppler shift, of every returning light signal enables LiDAR sensors to collect critical data for each point in the FoV. With per-point Doppler, the LiDAR will be able to determine whether a moving pedestrian will walk in front of your vehicle or safely head in the opposite direction.
However, while all LiDAR sensors – from pulsed-time-of-flight to flash and spinners – will experience this doppler effect, most are unable to measure it. Traditional LiDAR sensors only measure the amplitude of the return light (known as direct detection) and fail to measure other optical properties, such as phase and frequency, that are required to detect this Doppler shift. That’s because in order to measure that slight shift in frequency, the system must use a detection method called homodyne detection. This requires a fundamentally different system design to the legacy direct detection method, meaning that the design of most LiDAR systems on the market today are incapable of measuring Doppler, no matter how hard they try.
As there are clear advantages to measuring instantaneous velocity, some LiDAR companies have turned to Frequency Modulated Continuous Waves (FMCW) technology, which utilizes homodyne detection to enable the measurements of both the range and Doppler shift. However, despite common assumptions, homodyne detection isn’t exclusive to FMCW, which comes with its own set of trade-offs, such as speckle and pointcloud blur and has significant challenges meeting the points per second and resolution requirements needed to enable L4 autonomy.
At Baraja, our LiDAR combines our patented Spectrum-Scan™ steering technology and unique ranging technique, Random Modulation Continuous Wave (RMCW) paired with homodyne detection, to enable a high-performance Doppler LiDAR without any of the other issues often found in other new generation LiDAR designs.
Back to the benefits of homodyne detection: one of its key attributes, aside from increased detection range, is its ability to detect the velocity of every point instantly. It is often referred to as an instantaneous velocity measurement because it is able to measure the distance and velocity with a single measurement, versus the “brute force” approach that is required when using legacy LiDAR technologies that rely on “tracking” objects through multiple continuous frames and time consuming back calculations to determine an object’s velocity.
With homodyne detection, you can accurately and quickly measure the distance and velocity of every point in the pointcloud, which enables fast and reliable object detection and tracking of critical items in the FoV (pedestrians, other vehicles, animals, etc.) without the cost of heavy computations on the already strained perception algorithms.
Doppler measurement also increases the accuracy and reliability of segmentation. For example, when a group of points all have the same velocity, it is fair to estimate that the points are all coming from the same object. This segmentation process makes it easier to identify and separate individual objects while also reducing the likelihood of false object detection, thus, increasing the overall reliability of the data. In turn, improved reliability enables the perception stack to make quicker decisions, as there is an increased confidence to act on the data it receives from the LiDAR.
Finally, Doppler can detect and catch critical edge cases, an extremely important and potentially life-saving benefit necessary for L4 autonomous systems. Even if the chances are low for an event to occur, autonomous vehicles (AVs) must be able to swiftly and accurately understand and react to extreme scenarios. Without this capability, AVs risk the possibility of severe, potentially fatal accidents.
For example, say you have a group of pedestrians walking along the side of the road up ahead. Suddenly, one individual plunges into traffic without looking. Conventional systems wouldn’t be able to differentiate each individual at long distances, making its frame-to-frame “tracking” completely ineffective. With homodyne detection, which enables Doppler, the LiDAR system is able to instantly measure the velocity of the small cluster of moving points from the larger group, enabling the perception algorithm to identify and segment the individual darting into traffic. It can then provide that information to the perception stack, allowing it to make a quick and accurate reaction. This is just one example of many edge cases that Doppler-enabled LiDAR can solve.
Doppler isn’t the only thing that makes an effective long-range LiDAR. There are many other parameters that go into enabling this future.
At Baraja, we have combined our proprietary Spectrum-Scan™ steering technology with our unique RMCW ranging technique to develop a LiDAR capable of incredible resolution and precision with a fully solid-state fast axis and complete interference immunity.
These steering and ranging technologies alone add significant benefits to our LiDAR that set us far ahead of other technologies. However, Baraja has also invented a way to combine these advantages with a homodyne detection method to enable Doppler measurement. Prior to Baraja’s revolutionary LiDAR, homodyne detection was thought to only be capable with FMCW technology, which has a reputation for poor pointcloud quality and an inability to meet the maturity and reliability requirements for automotive applications.
Baraja is the only Doppler-enabled long-range LiDAR capable of powering autonomous vehicles without any trade-offs. With our LiDAR, you can expect exceptional performance and data that you can trust in any automotive scenario. Read our in-depth white paper to learn exactly how our LiDAR sensor works, or contact our sales team today to learn more.