The commercialization process of L3 and above autonomous driving is entering the "fast lane".
On the one hand, relevant regulations and standards for high-level autonomous driving have been promulgated continuously, injecting a "stimulant" into the development of the autonomous driving industry.
For example, Xin Guobin, vice minister of the Ministry of Industry and Information Technology, once said that he will launch pilot projects for the access and on-road traffic of intelligent networked vehicles, organize and carry out city-level "vehicle-road-cloud integration" demonstration applications, and will support L3 and higher-level automatic vehicles. Commercial application of driving functions.
At the same time, the Ministry of Industry and Information Technology will also release a new version of the "Intelligent Networked Vehicle Standard System Guide" to speed up the formulation of more than a dozen key and urgently needed standards, including supporting the commercial application of L3 and above autonomous driving functions.
On the other hand, major car companies are also accelerating the development and application of advanced autonomous driving technologies. At present, Xpeng, Ideal, Avita, Zhiji, Huawei Wenjie and other brand models are planning to upgrade from high-speed NOA to urban NOA. This is also seen as a critical step from L2 to L3.
It is foreseeable that with further policy support, China's high-end autonomous driving industry is about to usher in rapid development, and automotive software tracks such as OS and middleware will also enter a period of rapid development. Behind this, whether it is intelligent driving, intelligent cockpit, body control or information network security, etc., it is inseparable from the empowerment of software.
ArcherMind has been deeply involved in the field of smart car operating systems for more than ten years, and has reached in-depth cooperation with world-renowned chip manufacturers such as Qualcomm, Renesas, NVIDIA, and Xinchi, and has established a joint venture company BMW ArcherMind with BMW. In order to further develop the research and development business of smart car operating systems and middleware products, ArcherMind Technology also established a subsidiary company - Zhida Chengyuan. At present, relying on rich operating system software development experience and underlying chip capabilities, the subsidiary has launched a fully self-developed smart car operating system "Fengsheng" FusionOS.
Generally speaking, facing the needs and development trends of the autonomous driving industry, ArcherMind and Zhida Chengyuan have built a professional intelligent driving technology team, and launched the intelligent driving software system FusionDrive2.0, NVIDIA authorized automotive vision laboratory, intelligent FusionStudio, a driving simulation system, and other combined punches from systems to special projects, from software to tools, etc., to accelerate the creation of autopilot products that can be commercialized and implemented by car companies and first-tier suppliers.
Intelligent driving software system FusionDrive2.0, a bridge to mass production of autonomous driving
In the era of smart cars, autonomous driving systems need to collect, process and transmit a large amount of data in real time, and communication and data transmission between vehicles are becoming more and more important. Therefore, high-performance communication middleware, as a core component of the entire autonomous driving software architecture, will be a bridge to mass production of autonomous driving.
FusionDrive 2.0 is compatible with many mainstream car-level intelligent driving chip platforms such as Nvidia and Horizon. With real-time communication scheduling as the core combined with mass-produced FusionWise HPC software middleware, it can provide customers with highly reliable intelligent driving underlying software that meets functional safety requirements. system.
The basic middleware provided by FusionDrive 2.0 includes communication management, DLT, network management, persistence, routing management, diagnosis, big data management, calibration, time synchronization, power management, OTA, health management, functional safety, information security and other functional modules , and by extending support for protocols such as SOME/IP, DDS, and TSN, it provides more efficient, reliable, and secure data interaction and collaborative control capabilities for intelligent driving.
NVIDIA authorizes the automotive vision laboratory to create the most complete scene and material library in China
At present, the automotive perception system is showing a trend of cross-integration of various technical routes. Chip manufacturers are trying to help customers achieve full coverage of various roads and traffic scenarios through the improvement of soft power, and give full play to the camera under dynamic and complex driving conditions. performance. Because the iteration of ISP hardware requires long-term technical investment, and the improvement of soft power is more in line with the current industry status.
As an NVIDIA ecosystem software partner, ArcherMind has deep accumulation in visual software and ISP and other technical fields. In 2022, it will join hands with NVIDIA to jointly establish an automotive imaging laboratory, which is dedicated to the laboratory's in-vehicle depth vision capability building, providing intelligent driving Relevant companies provide intelligent driving vision solutions based on Orin and Xavier chips.
At present, the automotive vision laboratory authorized by NVIDIA has established 1,000+ standardized automotive vision scenes and 80,000+ material libraries, and is committed to creating the most complete scene and material library in China. The automotive vision laboratory builds a standardized and streamlined data processing pipeline for diverse solutions, and optimizes sensor parameters to improve algorithm performance. In addition, the laboratory also has special instruments and software tools developed for autonomous driving road tests to solve the industry problem of long-tail effect of commercial landing data for autonomous driving.
The ISP takes on a very important task and determines the performance "ceiling" of the back-end image processing. ArcherMind Technology and Zhida Chengyuan provide mass-production-level ISP services based on NVIDIA DRIVE Orin and Xavier system-level chip products for autonomous driving-related companies, and have good cooperation with global mainstream sensor manufacturers OmniVision, ON Semiconductor, Sony, etc. relation. At present, it has obtained commercial appointments for a number of intelligent driving projects, including domestic top traditional car manufacturers and new force car manufacturers, and products and solutions equipped with relevant software platforms have been put into mass production.
FusionStudio, an intelligent driving simulation tool, completes the last mile of mass production of autonomous driving
Before mass production, self-driving cars need to go through a lot of road tests to meet commercial requirements. Efficient closed-loop data flow is the key to the implementation of autonomous driving research and development. However, the cost of road testing is too high, and open road testing is still restricted by regulations. It is difficult to reproduce extreme traffic conditions and scenarios, and there are hidden dangers in testing safety. Therefore, virtual simulation, as a technology that can significantly reduce the cost, time and risk of road testing, while improving testing efficiency and accuracy, has become the mainstream choice, and this technology can be applied to the entire V-shaped process to ensure product quality.
FusionStudio is a suite of simulation tools for the FusionDrive system that helps developers and testers improve the efficiency of automotive software development and integration. FusionStudio and the intelligent computing platform realize interoperability through high-speed Ethernet, and are compatible with various mainstream hardware platforms.
In terms of high precision, FusionStudio uses an advanced intelligent driving software platform to build a physically accurate simulation platform, which can quickly and efficiently conduct large-scale autonomous vehicle testing and algorithm verification.
In terms of complex scene reproduction, developers can create diverse synthetic datasets for rare and complex scenes, including physically-based sensor data and pixel-accurate ground truth labels. These labels include depth, velocity, occlusion and other parameters that are difficult to label;
In terms of closed-loop simulation, a virtual proving ground for autonomous vehicle development can be created by combining physically accurate simulation with a high-fidelity 3D environment. Closed-loop testing can be performed on individual software components or the entire autonomous driving stack;
In terms of scalable programmable API, users can customize the simulator according to their own needs through the Python interface, and easily build sensor models, vehicle dynamics, traffic model extensions, or custom hardware interfaces.
Overall, the commercialization and popularization of autonomous driving technology needs to go through multiple stages, including technology research and development, scenario verification, policy support and market promotion. ArcherMind and Zhida Chengyuan will continue to strengthen investment in autonomous driving technology and research and development, provide more advanced and reliable autonomous driving solutions, help car companies and first-tier suppliers improve product competitiveness and market share, and jointly promote the industrialization of autonomous driving technology.