Renesas Electronics announced that it will work with startup Dibotics to develop automotive-grade embedded lidar processing

For autonomous driving technology, real-time and efficient implementation of environmental awareness is the first step, and the vehicle needs to know "Where am I?" To achieve environmental awareness in real time, and to use sensor fusion, these new requirements for sensors and processing chips. Lidar's 3D point cloud data provides more accurate and real-time data than cameras and millimeter-wave radars.

About half a month ago, Renesas Electronics announced a news that it would work with startup DiboTIcs to develop a car-level embedded laser radar processor that would eventually be used in ADAS and autonomous driving. The goal of the program is to enable OEMs, Tier 1 suppliers and Lidar manufacturers to develop real-time 3D mapping systems with low-power Lidar processors that meet vehicle-level safety and working environment requirements. .

Simply put, the two companies will combine their respective advantages, namely Risa's R-car series chip and DiboTIcs' 3D SLMA technology, and finally launch a SLAM on Chip solution. The so-called SLAM on Chip, which is to deploy 3D SLAM technology on a SoC chip, generally, to achieve high-performance real-time SLAM, now need to be completed on a PC, so if the two can implement this program Level and mass production, then good news for laser radar manufacturers and companies using lidar.

We know that for autonomous driving technology, real-time and efficient implementation of environmental awareness is the first step, and the vehicle needs to know "Where am I?" To achieve environmental awareness in real time, and to use sensor fusion, these new requirements for sensors and processing chips. The reason why laser radar plays a very important role in autonomous driving is that the laser radar's 3D point cloud data can provide more accurate and real-time data than the camera and millimeter wave radar. However, laser radars must be accurate and must rely on large amounts of data. It has been previously reported that the products currently being developed have a maximum of 300 lines, so the processing of these data requires higher processing power requirements.

The cooperation between Renesas and DiboTIcs is to enable the 3D SLAM technology relying on laser radar to be completed on one chip. In this collaboration, Renesas offers hardware solutions, while DiboTIcs provides software processing solutions. Different from the traditional SLAM technology, according to the introduction, the SLAM technology that the two companies cooperate will only use the 3D data of the laser radar, without additional data access such as inertial navigation, GPS, wheel speed sensor and the like. Therefore, the amount of data input can be reduced and the computing power requirement can be reduced.

For Renesas, this is a matter of increasing the competitiveness of its self-driving open platform Renesas Autonomy Platform. Previously, the car cloud reported that this platform was officially launched by Renesas this year. From the architecture announced at the time, Renesas did not put the laser radar into the camera, but for the camera and laser radar, the core is also An image processing chip of the R-car series.

Since Renesas released this platform, it has not announced the OEMs that will cooperate with Renesas on this platform. At the end of October this year, Renesas and Toyota, Denso's cooperation, only mentioned that Toyota and Denso will use Renesas' two chips: R-car series and RH850. Although these two chips are also used by Renesas on the Renasas Autonomy Platform platform, this also fully illustrates that Renesas' previous deep accumulation in the car is reflected in the on-board microprocessor and image processing. In the cooperation between the three companies, Denso is the ECU that provides Toyota with ADAS.

Therefore, if Renesas really wants to promote the Renasas Autonomy Platform, it needs to continue to increase the weight. Obviously, a lidar processing chip that can meet the requirements of the car class is very suitable for this "weight".

Dibotics CEO and co-founder Raul Bravo said they will provide OEMs, Tier 1 suppliers and laser radar manufacturers with a solution that will reduce development costs and accelerate product marketability.

Dibotics was founded in Paris in 2015. Its core competence lies in SLAM technology, including on different sensors such as laser radar, 3D camera, millimeter wave radar and so on. As mentioned earlier, one of the features of the SLAM algorithm used by Dobitics is the use of a single sensor, which means that the lidar uses only lidar data, the camera only uses the camera's data, and does not need other signals, such as odometers, habits. Guide or a variety of different sensor fusion, but also can ensure long-distance without offset.

From the Lidar, Dibotics has an enhanced laser radar (Augmented LiDAR) software. This software has two features:

1. Automatic point cloud classification technology, the software can automatically classify the point cloud data collected by the lidar without using deep learning, no need to have experience accumulation or comparison with map data, etc., detection and tracking movement Object shape, velocity and trajectory

2. Data fusion between multiple lidars can be achieved without calibration or synchronization

In July last year, Velodyne announced that it would cooperate with Dibotics on SLAM technology. The two companies will jointly develop SLAM technology based on Velodyne's laser radar.