Did NVIDIA's three-gun autopilot industry start to tilt?

Probe domestic switch needle KG-300K needle head diameter is 3.0mm normally open switch needle

Modern skhynix emmc 8G industrial grade H26M31001HPR

Industrial Router Crystal 3.2*2.5mm 3225 26M (26.000MHZ) 12PF 10PPM 20PPM 30PPM

100V, 250mA, surface mount, high speed switching diode

Not long ago, the ambition of NVIDIA ushered in five new partners. It seems that they have announced that the rhetoric of going straight to mass production during the R&D phase is indeed very attractive.

Subsequently, Volvo and Autoliv bundled NVIDIA chariot, they must rely Drive PX 2 platform in 2021 production autonomous vehicles. In addition, Tier 1 and Hella, the Tier 1 suppliers, have joined in, and the two companies will work with NVIDIA to promote the development of new car safety assessment test standards for the Autonomous Era era, helping large-scale deployment of autonomous vehicles.

However, these are not the most important partners, the German giant Volkswagen is the biggest fish NVIDIA has caught. In the future, the two companies will cooperate in depth in the field of AI and deep learning, so that these two technologies will shine in more fields.

Prior to this, the lineup of car manufacturers and Tier 1 suppliers around NVIDIA has been quite strong, with Tesla's commercial vehicles already deployed the Drive PX platform, and Audi is preparing to launch a platform based platform in 2020. 4 Self-driving vehicles.

Another super giant, Toyota, is preparing to drive its own ADAS system with NVIDIA's platform. In addition, well-known companies including Daimler and Bosch are firmly on the side of Nvidia.

* As can be seen from the above diagram, the Nvidia camp is the most powerful.

Danny Shapiro, senior director of the NVIDIA Automotive Division, once said: “The self-driving car is gaining momentum.” Now our main task has shifted from the development phase to the mass production phase.

Meanwhile Shapiro also mentioned that there are already 225 companies to join the Drive PX platform, these companies perfectly cover the entire middle and lower reaches of the automotive industry, the automobile manufacturers, Tier One suppliers, truck manufacturers, high-precision mapping company sensor And all kinds of new startups have been exhausted.

When it comes to working with the public, Shapiro said that AI is not vehicle-specific now, it will also become a back-end system for the data center, providing more traffic patterns, traffic and driving habits, and understanding the entire transportation ecosystem.

IHS Markit car agrees with the electronics sector, principal analyst Luca De Ambroggi. “NVIDIA is indeed powerful, it not only has its own solutions in cutting-edge technology, but also does not fall under the foundation. After all, for many car manufacturers, basic functions like predictive diagnosis, maintenance, network security and traffic management are It is really just needed."

Previously, many skeptics thought that NVIDIA's AI platform was only a laboratory product, and it was impossible to use it in a production car. However, NVIDIA used the actual action to frequently respond to this wave of singers.

Ian Riches Strategy Analytics company executives said in an interview: "From the point of view of a series of public conferences, Nvidia now really in the lead and do not forget this is the judgment prior to a series of major good news, five new partners. The arrival is even more determined by my judgment."

Phil Magney, founder of VSI, agrees. “Now there are fewer and fewer car manufacturers that don’t use the NVIDIA Drive PX platform. Of course, this doesn’t mean they can all go into mass production quickly. After all, they don’t want to be mice, don’t want to pay for the safety of AI and automation technology. ."

Are these cooperations strong enough?

It is worth noting that the five new partners signed with NVIDIA are non-exclusive agreements, and that more than a few companies, chip suppliers, Tier 1 suppliers, etc. are needed to design highly automated vehicles. Roles such as car manufacturers and software developers are critical.

Therefore, the biggest question is whether these cooperation agreements are stable? How likely are these automakers to convert R&D platforms such as Intel?

Riches of Strategy Analytics believes that “now focusing on one platform does not mean that the future will not change, but at least it proves that this solution is the best choice in the short to medium term.”

“Replacing the platform is costly. Software development is optimized for a particular platform, and engineers have a familiarity with development tools,” he added.

VSI's Magney said: "The autopilot industry does not have the concept of plug and play. Once a car manufacturer has made a choice, it may continue to stick to it. Very small."

At the moment, VSI has also devoted itself to the research and development of self-driving cars for research purposes. Magney said: "The development of the autopilot function is very difficult. We need to string together the basic code, complete the synchronization of the sensor, calibrate the torque signal and control the delay problem, which means huge engineering resources."

“In addition, developing and integrating software into the hardware platform is a time-consuming and laborious task. Although the abstraction layer of some autopilot stacks makes it easier to integrate, there are still various insurmountable gaps in development.” He added.

IHS Markit's Ambroggi believes that the relationships between companies are subtle and complex. He pointed out: "If the design of autonomous vehicles is divided into two parts, hardware and software, it is obvious that NVIDIA hopes to sell the complete solution directly to the customer. But in fact, I am afraid that the end is not the same."

De Ambroggi lists two of the most probable scenarios. The first is that the car manufacturer and the Tier 1 supplier fully accept Intel's solution, the other is to prioritize the software and algorithm parts, and the hardware to choose according to the needs. Such a solution allows automakers to bargain for SoC prices, while not being completely covered by NVIDIA.

But in any case, the high added value of the software can make NVIDIA earn a lot of money.

De Ambroggi believes that Intel's capabilities in the data center and training infrastructure will be on par with Intel in the future. “As a result, it will be able to offer customers a wide range of optimal solutions including IT infrastructure in the future.”

He pointed out that the fundamental reason for truly complicating the platform is that AI needs to extract data from different sensors from different vendors and push the processed information to the vehicle ECU, which is usually from different suppliers.

"As long as the performance to meet the requirements of OEM's, AI in the end what the data would become less important." De Ambroggi concluded. “Obviously there needs to be a bit of standardization here. I think both automakers and Tier 1 suppliers should make early plans.”

Cooperation reshuffle

As the whole industry gradually transforms from the R&D vector production stage, the cooperation relationship between various manufacturers has begun to undergo a new round of reshuffle. After Volvo and Intel work together, the AImotive that originally provided software for it has been smashed into the cold. Replaced by the new superior Zenuity.

*Volvo auto driving concept car interior

The latter is a joint venture between Volvo and Autoliv with the main goal of developing the next generation of autonomous vehicle technology. Zenuity will provide Volvo with autopilot software, which will be sold to third-party automakers in the future.

Simple migration path

Shapiro calls the process of developing a vector transformation called a "simple migration path" because the entire process does not need to be reconfigured and the system code is perfectly compatible.

In Shapiro's view, NVIDIA's strength lies in the "openness" of its architecture. He called his architecture "a software-defined, AI-driven open autopilot platform."

It is understood that the openness of the system allows Tier 1 suppliers to customize new software, add various functions and complete upgrades at regular intervals. NVIDIA's chip architecture can run many different software. "It's quite open and highly scalable, helping automakers stay at the forefront of technology," Shapiro explained.

Magney believes that NVIDIA has completed the construction of a complete autonomous car stack. “It has hardware, open tools, and a large file base in it, and Tier 1 suppliers and other third-party developers can build their own applications. In addition, it is ready for mass production. ADAS and autonomous driving solutions."

Other manufacturers have no way to fight back?

De Ambroggi believes that NVIDIA has an unparalleled advantage in the AI ​​field. It has a large number of partners and has formed a cluster advantage.

In the hardware market, NVIDIA and Intel are both in the world, but De Ambroggi believes that the future situation may change, because car manufacturers may prefer a hardware-independent platform.

“Know that the AI ​​platform is not the only electronic system that controls autonomous vehicles.” From the perspective of ADAS/autopilot (especially in the short to medium term), the entire industry may reshuffle. “Therefore, traditional electronics manufacturers and suppliers may have a place in the future. It takes more effort to integrate security and AI, and we also need some system redundancy.”