Introducing in-house lidars for our self-driving vehicles
How our self-driving car sees the world is a critical element of our autonomous platform. We use the latest technology to give our cars a full understanding of their surroundings, ensuring that they can safely navigate any number of driving scenarios. Lidars are a critical component of our self-driving platform, using lasers to accurately detect objects around our vehicles. Today, we’re excited to announce that our engineers have begun testing prototypes of our own lidars. We’re currently testing our sensors for use in our self-driving cars in Moscow and expect to use them on our delivery robot as well.
We are working on two types of lidar. The first is a solid-state lidar, which has a 120 degree field of view and can, for example, provide a highly detailed view of objects in front of the vehicle. The other lidar provides a 360 degree view for the car, creating a detailed model of the car’s entire environment.
After almost three years and over 1.5 million miles of testing, we have a better understanding of exactly what kind of data we need from the sensors. To achieve better flexibility, we’ve created versatile software that allows us to tune the lidars’ scanning patterns while driving. The custom parameters enable our lidars to adapt to various driving conditions, like freeways, dense city streets, and diverse weather. For example, our lidar can focus on a specific object and adjust its scanning pattern to accurately determine whether an object is a pedestrian, bicycle, or something else at a distance of 200 meters.
“Third-party lidars analyze and filter data as soon as it’s collected. Using our lidars, we receive more information about the vehicle’s surroundings since we can access the sensors’ raw data,” says Dmitry Polishchuk, Head of Self-Driving Cars at Yandex. “With our lidars, we can analyze the raw data and synchronize it with information from other sensors, so that the car can better identify objects. In addition, our current prototypes are already half the cost of existing devices. With the transition to mass production, the cost of our lidars will be even lower, and we will ultimately save up to 75% on the cost of sensors.”
Lidars are one of the three essential sensors of our self-driving cars, along with cameras and radars. Cameras examine the vehicle’s surroundings, and determine not only the type of objects around the vehicle, but also understand details such as the words on a road sign and the instructions of traffic lights. Radar uses radio waves to determine the location and speed of objects around the car, though its accuracy is not high enough to understand the type of object. Lidars create a detailed 3D model of the vehicle’s environment from millions of reflections, by measuring the time it takes for the laser to bounce off objects. It does not depend on the lighting conditions as do cameras, and provides much more precise information about the objects and their location than radars. By combining data from these three types of sensors, our self-driving system gets the most detailed picture of the world around.
In addition to lidars, we are also designing our own camera. Thanks to the High Dynamic Range (HDR) of the camera, it can detect objects in low or extremely bright lighting simultaneously. It can also quickly adapt to sudden changes in lighting, such as entering a tunnel or leaving a parking garage. Our camera can also process LED lights without the flickering that often affects how cameras see LEDs. This is an important safety feature for self-driving applications as more traffic light bulbs are replaced with LEDs and its critical to detect them in all conditions for the vehicles to react appropriately.
The development of our own cameras and lidars ultimately enables greater control of the platform and will help us to continue to advance our self-driving cars. Our in-house lidars will also lower the cost of building self-driving vehicles, creating more applications for autonomous technology. We look forward to announcing more updates on the continued progress of our autonomous tech as we create safe, universal, and accessible transportation.