On December 16, the "World Intelligent Vehicle Conference" was held in Guangzhou. Liu Bin, co-founder of Juefei Technology, was invited to attend the conference and demonstrated Juefei Technology's "Autonomous Driving Full-stack Converged Computing Solution" to the guests.
From 2021 to 2030, autonomous driving has entered a golden decade of rapid development. "Lidar on the car" has become the latest label for automotive intelligence. According to public information, by 2030, 20%-25% of the vehicles in the market will have high-level assisted driving capabilities every year, which means that there will be 4-5 million standard spaces per year. In the face of such a huge market space, the competition for mass production ranking around high-end automatic driving has begun. Routes using lidar as sensors and fusing other perceptual data have been verified and deployed on a large scale in L4 scenarios such as Robotaxi and low-speed logistics. However, for passenger cars, it is a new issue. In terms of the productization of technologies such as fusion perception and computing, there is still a large technological gap in the entire industry.
In order to meet the problem of OEMs overcoming the mass production of high-level intelligent driving systems (NOA) for passenger cars, and at the same time cooperate with the general trend of passenger cars on lidar to achieve safe and reliable high-level intelligent driving, Juefei Technology has launched the "Autonomous Driving Full Stack Fusion Computing Solution", which fully meets the urgent needs of intelligent driving cars from "assembling" lidar to fully "playing" and "applying" lidar, and has won the favor of many OEMs.
This set of solutions can provide customized front-loading and rear-loading positioning algorithms, multi-sensor fusion perception algorithms and dynamic traffic information services for different scenarios and customers, to meet the needs of all stages of automatic driving, and to upgrade vehicles from low-level assisted driving to NOA empowerment.
Juefei Technology Liu Bin speech transcript (recommended to watch under wifi)
The solution can provide pure vision-based fusion perception and fusion positioning services for L2+ autonomous driving systems, and can adapt to mainstream visual SOCs and posture sensors.
For high-level autonomous driving, it is also possible to integrate high-performance platforms such as Huawei MDC, Horizon J5, and NVIDIA Orin, and finely deploy products and solutions such as fusion perception and fusion positioning according to different computing power needs.
The core capability of this program is the fusion perception technology of Juefei Technology, which is also the core technology route that Juefei has adhered to since its inception.
Non-convergent perception technology builds urban holographic intersections
The non-technological fusion perception algorithm fuses and calculates the raw data of vision, point cloud and millimeter wave, and then performs 3D tracking to output more accurate and complete perception results. At the same time, with the help of high-precision maps and global samples at the end of the road, the behavior trajectory of traffic participants was accurately predicted.
Through a large number of fusion perception ability practices, Juefei has developed a number of point cloud algorithms, which can enable larger models to achieve higher precision LiDAR + multi-channel vision fusion perception on the embedded controller of the mass production car, including environmental perception, situation awareness and behavior prediction. This year, Juefei Technology also launched the high-precision fusion positioning terminal system "Minghong" applied to the end of the autonomous driving car.
Its built-in environment-aware algorithm, FPGA vehicle-level computing chip and other technical capabilities can match the features of the high-precision map mounted on the car, and use the visual and feature matching results to provide key information input for inertial and satellite guidance. Especially in urban multi-path scenarios, multi-layer interchanges or tunnels, and other scenarios where the guard and inertial navigation are seriously disturbed, the accurate and stable output of vehicle positioning can be maintained.
Juefei technology high-precision fusion positioning technology architecture
"Since the beginning of this year, the computing power of the entire car is constantly improving, and the high robustness and plug-and-play characteristics of 'Minghong' can quickly help the OEMs complete the entire iterative process." Liu Bin said.
Not only that, Juefei Technology has built a data closed loop for the target perception and prediction of automatic driving through the fusion perception and roadside scene data model of the car end, and can provide full-stack technical services and empowerment for data self-updating.
At present, Juefei has applied this set of "automatic driving full-stack fusion computing solutions" to different scenarios, including vehicle lane changes, long tunnels, and under viaducts, and has reached cooperation with a number of OEMs to provide customized, high-specification, mass-produced product solutions and services for many partners from the dimensions of accuracy, reliability and cost.
In this conference, Juefei Technology also won the "Intelligent Vehicle Technology Innovation" award, which also means that Juefei's multi-sensor fusion perception technology has been further recognized by the market, and the automatic driving full-stack fusion solution will accelerate the mass production of passenger car NOA, bringing more possibilities for the deepening and development of high-end automatic driving and intelligent transportation to the consumer market.