Wen Hao Ott is driving? A
NVIDIA's start-up acceleration plan has become an important part of NVIDIA's development in the AI field.
65438+February 15, NVIDIA GTC China conference was held online. At the GTC conference, NVIDIA released a faster AI chip to build the world's first "intelligent distribution city" with JDL Jingdong Logistics, and the world's first generation of end-to-end network solution with 400Gb/s network speed, NVDIAMELANOX 400 G? Unlimited bandwidth
As an important part of this GTC, NVIDIA presented 12 start-ups that stood out from more than 0/00 companies in NVIDIA.
The selected enterprises cover conversational artificial intelligence, smart medical care/retail, consumer Internet/industrial applications, deep learning applications/accelerated data science, autonomous machines /IOT/ industrial manufacturing, and self-driving cars.
In the field of self-driving cars, Tag Star and Hongjing House are the first to be recognized by NVIDIA and the industry, among which NVIDIA will provide technical, marketing and financing support.
As an independent startup company of full-stack unmanned transportation solution in mining area, Tagg Star realized multi-group operation of unmanned mining trucks as early as 20 19. This year, unmanned night work and unmanned cab were realized, and large-scale project deployment began.
In order to be able to quickly implement the actual technology, Tagg Zhixing relies on NVIDIA? Gjertsen. TX2i and NVIDIA? Gjertsen. AGX？ Xavier computing platform developed the unmanned transportation solution "Mine Valley" in open pit mine.
The mine valley includes vehicle-mounted system, ground system and cloud control system, Shu Tian. The system takes open pit mine as the main application scenario, and solves the problem of perceived pain points caused by high dust in mining areas and blurred road boundaries.
In addition, M-box, a self-developed on-board hardware computing platform, has completed the third iteration of a large version, integrated the 5G communication module, and passed a number of vehicle regulatory tests and certifications of China Institute of Environment, Reliability and EMC Metrology, making it the earliest batch of 5G integration in China. +? C-V2X communication, high performance parallel computing, high safety decision control, vehicle domain controller certified by vehicle regulations.
As a start-up company, TAG Star signed a commercial contract of over 100 million yuan in 20 19. Among them, the mine truck driverless project signed with Baiyun Obo Iron Mine of Baotou Steel Group completed the second phase acceptance on 10 this year.
The unmanned project contract signed with the State Power Investment South Open-pit Coal Mine is the first unmanned transportation project in China, which was delivered for acceptance in June this year.
The contract for 200 unmanned wide-body vehicles in cooperation with EPC Central, the largest coal mine in China, has been delivered to three operation groups, and Yongshun Coal Mine in Ordos, which cooperates with Central, has realized 24-hour continuous operation.
It is reported that the conversion rate of POC project to commercial orders exceeds 80%, and subsequent orders exceed 300 million yuan.
Another self-driving startup, Hongjing Zhijia, was founded on 20 18. The company mainly develops vehicle-level autopilot system solutions, and its main products are software and hardware integrated autopilot computing platform (ADCU? –? Autonomous? Driving? Calculation? Unit).
This is the only universal platform solution that supports advanced autopilot (L3 and L4) in China market at present. It integrates algorithm software, architecture design, vehicle safety and chip, which can achieve the purpose of efficient optimization, safety redundancy and controllable cost and power consumption, and can meet the needs of large-scale pre-assembly and mass production.
During the GTC meeting, Hongjing Home plans to cooperate with the main engine factory to mass-produce L 1\L2 self-driving passenger cars in the fourth quarter of 2002/KLOC-0, providing advanced driving assistance functions including automatic parking, AEB (automatic emergency braking) and ACC (adaptive cruise control).
At the same time, Hongjing House adopted NVIDIA in its L3+ advanced automatic assisted driving system. Xavier? GPU has landed in the fields of high-speed trunk logistics, intelligent bus and travel.
Hongjing Zhijia is also cooperating with several mainstream commercial vehicle factories and technology companies to develop L3-class self-driving trucks to help logistics companies save manpower, reduce fuel consumption and improve driving safety. According to the plan, its L3 truck can save 5% fuel, reduce the labor cost by half, and greatly reduce the accident rate.
Greg, Global Vice President, Enterprise Market and Developer Program, NVIDIA? Estes said in a speech, "These can quickly land in the field of autonomous driving, and also benefit from NVIDIA's support in investment, marketing and technology."
First of all, NVIDIA allows them to contact the technical resources inside NVIDIA, and they can establish personal contact with NVIDIA, which is very important for startups. Communicating with technicians is far better than consulting documents.
Second, startups want to expand their scale. We know that many startups start with local GPUs and then want to expand to the cloud.
So NVIDIA can provide them with corresponding programs, which greatly improves the usability of GPU. NVIDIA offers subsidy programs and preferential prices to help them get started and gradually mature.
Finally, a very important item is popularity.
NVIDIA helped them with marketing plans, established successful cases with startups, and relied on NVIDIA's marketing resources to enhance the visibility of startups.
Greg. Estes said, "In many cases, part of the job is to communicate with venture capital institutions. NVIDIA has built a bridge between venture capital institutions and startups, and once again raised the visibility of startups. "
In terms of autonomous driving technology, Invista has also made innovations.
First of all, in the end-to-end data collection of autonomous driving, NVIDIA will collect data through a large number of sensor devices, while in the large DGX? The data center training model on SuperPOD generates a trained neural network model for deployment in automobiles.
Before deploying these models to cars, NVIDIA should test them through hardware-in-the-loop simulation.
NVIDIA will use the actual AI hardware in the car running these models to simulate and synthesize the information seen by these models, including sensing components, planning components and forecasting components, which can use all sensors to perceive the surrounding environment of the car.
These components can predict the behavior of other cars, pedestrians and other traffic participants in the scene, so as to plan ahead according to the possible behavior patterns of other traffic participants.
In order to handle the workload of different levels of autonomous driving, NVIDIA will adopt various products and solutions based on Ampere architecture specially built for edge applications.
For the assisted driving function, we can use the embedded chip based on Ampere architecture to provide 10 trillion operations per second, and the energy consumption for handling this task is only 5 watts.
For L2 autopilot, because it can provide 45 watts of energy consumption and 200 watts per second? Topsy Olin handles the workload.
For a fully self-driving L5 driverless taxi, NVIDIA will generate a computing power per watt? 100TOPS？ Products.
To sum up briefly, enterprises that have been demonstrated by start-ups in NVIDIA can obtain the capabilities provided by NVIDIA through the most direct channels, which are not limited to AI technology, and are cost-effective.
Greg. Estes revealed in an interview that "the NVIDIA project has covered nearly 7,000 AI start-ups around the world. NVIDIA hopes to help start-ups develop at the critical stage of product development, prototype design and deployment through the start-up demonstration in NVIDIA, and each member enterprise can continue to obtain tailor-made assistance rights, which provides a basic tool for the development of start-ups. "
At present, autonomous driving technology is in a period of rapid development. After financing and technology integration in the first half of the year, autonomous driving will definitely enter the stage of technology landing in the second half of the year. How do start-ups break through in the new round of technological explosion? Besides excellent technical strength, they also need mature AI companies like NVIDIA to empower them.
This article comes from car home, the author of the car manufacturer, and does not represent car home's position.