Steering Autonomous Vehicles to Level 5
Michael Alba posted on September 23, 2020 |
Automotive expert Nand Kochhar explains the design considerations of autonomous vehicles.
Siemens Digital Industries Software has sponsored this post.
Autonomous vehicles use many different sensors to understand their environment. (Image courtesy of Siemens.)
Autonomous vehicles use many different sensors to understand their environment. (Image courtesy of Siemens.)

Cars are becoming smarter. Someday—hopefully soon—they will be able to drive themselves, achieving the Society of Automotive Engineers’ fabled fifth level of autonomy. There remains a lot of road between here and there, but engineers have their feet firmly on the accelerator.

Or maybe that should be Xcelerator. That’s how Nand Kochhar, Vice President of Automotive and Transportation Industry at Siemens Digital Industries Software, might put it—Xcelerator being the apposite name of Siemens’ product development portfolio. A thirty-year veteran of the automotive industry, Kochhar spends his time developing the tools that help engineers get closer and closer to Level 5.

Nand Kochhar, Vice President, Automotive and Transportation Industry at Siemens Digital Industries Software. (Image courtesy of Nand Kochhar.)

Nand Kochhar, Vice President, Automotive and Transportation Industry at Siemens Digital Industries Software. (Image courtesy of Nand Kochhar.)

Engineering.com sat down with Kochhar—virtually, of course—to learn about how vehicles are becoming autonomous and what’s needed to get them there.

Engineering.com: What’s the current state of autonomous vehicles?

Nand Kochhar: There are the SAE [Society of Automotive Engineers] Levels 1 through 5, and Level 1 and 2 are pretty much adopted by the industry for most OEMs. Level 3 is the cutting-edge technology, and companies are already going into Level 3 autonomy.

We are headed towards Level 4 over the next few years. And several industries, especially trucking, are already leapfrogging to the next level of autonomy. For example, Waymo is starting to test Level 4 vehicles in Texas. I think technologies are maturing fast and we are headed towards those levels of autonomy. The trucking business might be taking it to the next level of autonomy sooner than the automobile itself.

SAE levels of driving automation. (Image courtesy of National Highway Traffic Safety Administration.)
SAE levels of driving automation. (Image courtesy of National Highway Traffic Safety Administration.)

How do you model an autonomous vehicle system?

We could divvy it up into three major areas with an analogy of the human body. The first part is perception. The function provided by the human eye is being performed by cameras, LiDARs and radars. So that's the perception algorithm.

After you’ve perceived the images, you bring them back for decision-making. You get this fusion of different images happening at a chip level. From a physical standpoint, it's the smallest component. But from a decision-making standpoint, you are starting to do the fusion as well as the processing of the information—an analogy for human brain function.

The third part is sending that message back to the mechanical systems to take some action.  In our human body analogy, this would be the function of the muscles. As an example, for emergency braking or adaptive cruise control, you need to send those messages based on perceptions from the signals coming in to the braking system. The equivalent of that would be a person pressing on the brakes. The controls send the message to the brake system and apply those brakes so you can come to a stop in an emergency.

(Image courtesy of Siemens.)
(Image courtesy of Siemens.)

You need a full vehicle model which is capable of doing all the vehicle dynamics and the chassis development and the controls. That's where you will see a combination of all these technologies coming to deliver the intended function—in this case, stopping of the vehicle.

How do developers approach decision-making?

The technology is continuing to mature. The decision-making level could happen at a chip or SoC [system-on-a-chip] at a board level, instead of sending the data back to a central unit for processing and then making that decision. Different companies have different strategies to implement these decisions. They pick what extent they need to engage at the chip level and at a centralized computer level.

I think the boundaries on automotive are going into the semiconductor market. Look at some of the more advanced chips; for example, NVIDIA chips versus the traditional. The trend is that these decisions that happen at a chip level will speed up the processes and address the issue of the latency.

But that's where I think these two industries, the semiconductor industry and the automotive industry, are coming together. They’re making waves about how the automobile is a computer on wheels.

What is Siemens doing for autonomous vehicles?

Our Xcelerator portfolio is very broad. It covers aspects from design, simulation, development to manufacturing. Siemens is a trusted partner for all major OEMs and Tier suppliers in the automotive and transportation industry. The majority of automotive OEMs will use some elements of our portfolio, including the U.S. big three—Ford, GM, and Chrysler all work with Siemens at different levels.

In the ADAS [advanced driver assistance system] arena, our product offerings include what we call our chip to city initiatives. If you look at the product development and manufacturing cycle, we have pieces used through the entire chain. That goes all the way to chip design, graphics, simulation and testing, and the highest level will be a vehicle in an environment. For example, creating different drive scenarios using a product like Prescan360, which lets you do a lot of the sensor development and learn how that development tracks with traffic and city infrastructure. That’s an example of usage of our portfolio.

How big is the cost barrier to autonomous vehicles?

Cost in the automotive industry is always a challenge. When new technologies are coming in, cost is a challenge until the volumes grow and mass production is possible. Then the cost curve continues to come down with the economies of scale.

I look at the technology coming into automobiles in a way that is no different than any of the other technologies coming in—they come in at a very high cost. You can take the example of LED TVs coming out initially at a high price point, to the way they are today. Or computers, where they were. IBM were the first ones to lower the cost of those computers.

I think the technologies for all different levels of autonomous vehicles are under a similar curve. The cost is high, but on the other hand, consumers are seeing the value in those costs and the trends are adapting.

How do you convince the public that autonomous vehicles are safe?

As you know, safety is absolutely critical. That's why people are very careful taking baby steps before declaring a victory on any of this.

Of course, there's the growing pains of adapting to the new way of transportation, and that comes with its own concerns. Change is not easy for us as human beings to adapt. Like the adoption of any new technologies, it's not overnight, and it doesn't happen to 100 percent of the population at the same speed. Media coverage and educating the consumer are the big factors which give that perception of how the industry is progressing, and how these things are safe for consumers from a day to day perspective.

Regulations and certification will also play a role in delivering a safe product. The NHTSA [National Highway Traffic Safety Administration] in the U.S., for example. In other countries, the local authorized bodies have test standards to which they measure vehicles—not only for delivering safety, but also giving the stars-based safety rating on the vehicles. To the best of my knowledge, those standards don't yet exist for autonomous vehicles.

That's where the regulatory bodies come into play. There is some activity with Euro NCAP [the European New Car Assessment Programme], which we at Siemens are engaged in, looking into those test standards and even ultimately moving into what the regulations might be related to autonomous vehicles.

What are the biggest challenges facing autonomous vehicles?

Perceiving under the conditions of extreme snow, fog or rain—these things become more and more difficult, typically from the lane markings being affected. Different radar technologies are coming which will address those challenges. Nighttime driving can also be challenging, so there are different infrared technologies coming which will address those scenarios as well.

(Image courtesy of Siemens.)
(Image courtesy of Siemens.)

One very important thing is that in order to get equal miles to feel safe, you have to drive for millions of miles to test out all the possible scenarios. Scenarios of road conditions and other drivers, as well as the infrastructure. There are endless combinations of those scenarios, which you really can't reproduce in a physical test environment, and sometimes it's not safe to test those scenarios.

All that leads to virtual verification and validation as a key to the autonomous rate of delivery. Some of those edge cases in these scenarios can be produced and analyzed. You can play millions of scenarios and continue to run them to get effective decision making. I think virtual verification and validation is the absolutely necessary thing to do.

Are all autonomous vehicle developers using simulation?

The simulations have been pushed by the companies who are creating autonomous vehicles. Companies such as Waymo and Tesla, who are at the forefront of autonomous technology in the U.S., and globally many more players are pushing the envelope on the simulation front. Siemens plays a big role as a tool provider and supports them to make it happen.

Simulation is such a subject. I've spent 30 years in it, and it depends on the decision makers at the top of the house, so to speak. Some believe in simulation and they push for it; we don't have to sell anything because they tell me the same story: simulation is the only way. Others are mixed, always trying to justify the investment, and it takes a while to get that messaging.

This fits into the bigger story of digital transformation in general. It's a journey, not an overnight thing. Some companies get it and then they drive it within their organizations. Others need a lot of “proof in the pudding”, so to speak, and then they want to work with companies like Siemens and progress step by step. We are there to work with our customers at their pace.

When will we get to Level 5?

I would say it's far, without putting a specific year on it. Every day as the technology is maturing, there's a new player who comes on board with a timeline of when they're going to be able to deliver it. But for me as a consumer, as an engineer, I call that far.

How will we get there?

The real key for us is the process revolution going on. We have four key digital threads which enable these autonomous vehicles. One of the key digital threads is software and systems engineering, where we touch on a majority of the things we talked about. Then we have accelerated product development and smart manufacturing, where we touch on end-to-end concept to production. Then there is the digital thread on IoT and analytics, which connects and facilitates a lot of these closed loop processes for customers to monetize the benefits of all the other digital threads.

Xcelerator is a portfolio of products, services and the open platform for applications development. We can have a realistic, closed loop process, which enables comprehensive digital twins of design, manufacturing and service. Siemens has that realistic experience and practicality of doing these things, so we can help our customers carve their journey.


To learn more about Siemens Xcelerator, visit siemens.com/portfolio.

Recommended For You