Subscribe to this podcast

“With Apex.OS, AGCO gets a robust, reliable and already-certified base software for an end-to-end operating system, which allows AGCO to develop their autonomous applications much quicker. They can entirely focus on the application layer and bring autonomous applications currently being developed on their concept robot, Xaver, to market quicker and with more reliable results than before.”

— Jan Becker, Co-Founder and CEO, Apex.AI

Apex.AI, a company developing safety-certified software for mobility and autonomous applications, recently announced an expanded partnership with AGCO that will allow the OEM to add new capabilities to its autonomous Fendt Xaver concept vehicle. The row unit robot will incorporate Apex.OS, a software development kit that serves as base software for autonomous applications.

In this episode of the Precision Farming Dealer podcast, Apex.AI co-founder and CEO Jan Becker joins me to talk about the capabilities of Apex.OS, Apex.AI’s partnership with AGCO, Becker’s deep background in autonomous vehicle applications and much more.

The previous episode of the Precision Farming Dealer podcast featured an interview with Seth Crawford, AGCO’s senior vice president and general manager of precision ag and digital. Crawford talked about AGCO’s partnership with Apex.AI and the OEM’s plans for the base software. Listen to the episode here.

Full Transcript

Michaela Paukner:
Welcome to the latest episode of the Precision Farming Dealer Podcast. I'm Michaela Paukner, technology editor at Precision Farming Dealer. New episodes of this series are available wherever you get your podcasts. Be sure to subscribe to get an alert when upcoming episodes are released. If you listen to our last episode featuring Seth Crawford, AGCO senior vice president and general manager, precision ag and digital, you may remember him talking about AGCO's investment in and partnership with Apex.AI, a company developing safety certified software for mobility and autonomous applications. In June, Apex.AI announced an expanded partnership with AGCO that will allow the OEM to add new capabilities to its autonomous Fendt Xaver concept vehicle. The row unit robot will incorporate Apex.OS, a software development kit that serves as base software for autonomous applications. In this episode of the Precision Farming Dealer Podcast, Apex.AI co-founder and CEO, Jan Becker joins me to talk about the capabilities of Apex.OS, Apex.AI's partnership with AGCO, Jan's deep background in autonomous vehicle applications, and much more.

Jan Becker:
My name is Jan Becker. I am the CEO and one of the two co-founders of Apex.AI. Also, I actually, since 2010, lecture at Stanford University in the field of autonomous systems, vehicle automation and driver systems. The company's Apex.AI. We founded the company almost exactly five years ago in the summer of 2017 to build a robust and reliable base software. So, in a sense, a meta-operating system for both autonomous applications, but also mobility and vehicle applications overall. So, in a nutshell, you could call it an operating system for mobility.

Michaela Paukner:
How did you get the idea for Apex.AI? And walk through the company's growth since then.

Jan Becker:
I'll actually go back and walk you through my professional life, because that then also shows you why we are doing what we are doing and how it actually got started. So, after growing up in Germany and in the U.S., I did a PhD a little over 20 years ago, PhD work, building up a completely autonomous vehicle. And that was in the late nineties where that wasn't even a topic, that was seven years before the DARPA Urban Challenge, which really offset the trend and the revolution to build autonomous vehicles, so that was in the late nineties. And we back then automated endurance testing for Volkswagen, Volkswagen being one of the largest automakers worldwide.

Jan Becker:
It's a passenger car automaker. They operated a test track on which vehicles are tested, in this case, for one month, 24/7 driving on a specially designed test track to go through the entire mechanical life of a vehicle. So, it's extremely stressful for the vehicle, equally stressful for the test drivers and the intention was to automate it, which we then did in two years. So, we built what would today be a level four vehicle, be built in an entirely driverless vehicle. But in overall, vehicle technology wasn't as advanced 20 years ago as it is today. So, we first had to actually build a robot to sit on the driver's seat to then instead of a human driver, push the pedal, so the robot had three legs for three pedals.

Jan Becker:
We were driving manual transmission, had an arm for the gear shifter, another arm to turn the steering wheel and even more arms to operate all the levers behind the steering wheel. And that a robot ended up being more expensive than what the manufacturer, in this case Volkswagen, wanted to spend for the whole vehicle. Then I worked for Bosch a couple of years, Bosch being the largest automotive supplier worldwide. In the early 2000s, I developed driver assistance systems such as the traffic jam assist, which is very, very similar in functionality to what Tesla autopilot is today, except we developed it 10 years earlier.

Jan Becker:
But then automotive industry is very slow to adopt new innovation, so it took over 10 years to make that into a product, which then launched with Audi in 2015. We also build a left turn assist, which prevents you from making left turns into oncoming traffic, where you, as a driver, over or underestimate the velocity of the oncoming traffic, which then often leads to accidents. In 2006, I moved to Stanford University to follow [inaudible 00:04:47] in California to work at Stanford initially for one year, then one year became two year and now it's over 15 years. And initially, I was on the team that... I competed in the DARPA Urban Challenge in 2007, which was an event set up by DARPA the research agency of the department of defense with intention of showing that autonomous vehicles can actually be built and are possible. Was a series of three events in 2004, 2005 and 2007, and nobody successfully reached the goal of the first event in 2004. Stanford won the second event in 2005 and Carnegie Mellon won the event in where we came in second.

Jan Becker:
Then I started to build up teams at Bosch, again at Bosch in the research center in Palo Alto, one team in robotics and one team in autonomous driving. And in robotics, we then came in 2009 across a research company called Willow Garage back then, which was founded on Willow Road in the garage, hence the name Willow Garage. Which was really set up by one of the early Google employees, which was set up to make progress in autonomous systems and robotics. And they then quickly came to the understanding what the R&D community in autonomous systems really needs in order to collaborate better. Because it's a very hard problem, which is hard to solve for a whole community, but even harder to solve for an individual. And they concluded what the community needs is really collaboration.

Jan Becker:
And in order to foster and enhance collaboration, they started to build platforms. So, common platforms on which then the research community could collaborate upon, one being a robotic hardware platform, so more or less human not robot. And they then made 20 of those and gave away 11 for free to community members, and the community members included 10 universities and one corporate lab and that was my lab at Bosch. And then, they built up a software platform, which is called ROS, the robot operating system, which is open source and open source under a license, which explicitly allows use in commercial products. And the only condition for the recipients of those 11 robots was that all the results they would achieve in the next two years had to be contributed back into ROS, into open source, into the robot operating system. And this is really what jump start ROS in 2009, 2010.

Jan Becker:
Now, fast forward, 12 years later, 2022, ROS is really omnipresent in R&D for not just autonomous driving or robotics, but also for drones, for service robotics, for industrial robotics, for R&D. It's used in R&D for medical robotics, [inaudible 00:07:55], mining construction, and also, agriculture and farming, where we see a lot of companies use ROS for developing autonomous systems prototypes. It has a huge user community. So, we estimate over a million users, about 200,000 are registered, but since anybody can just download and use it for free, there're obviously more un-registered than registered users. Over a thousand robots in different vehicles are supported already out of the box, so it's by far the largest R&D community in the robotics space, in autonomous system space. Now we then use... We now then being at Bosch, we used ROS for a lot of prototype applications.

Jan Becker:
It actually helped us tremendously to accelerate the speed of prototyping applications. Early in my career, and I mentioned that before, systems took 10, 12 years from idea to go into a product. And the reason was simply that we as companies often redeveloped everything twice. First, we build a prototype, then we build a... Just to show the functionality. Then we worked on pre-development, that also took a couple of years and then, there was the third step, which was then really production development on an embedded platform, which then took another three, four years. And then became the result for innovation times of upwards 10, 12 years from idea to a product. Now, with ROS, the prototyping phase was accelerated tremendously, which we then saw already early on at Bosch.

Jan Becker:
So we, for instance, also built a robotic lawnmower, which we then put in consumer lawnmower, which we then put in a product. We also built an agricultural R&D robot through a public-funded project in Europe as a prototype, which... Also, their accelerated software development then tremendously. We then also put it into autonomous vehicles prototypes, but we also noticed what doesn't work at the time in ROS, which was really putting it into fast-moving, real time robots, where you need highest reliability, highest robustness, and ultimately, for certain applications or for certification. That then led to the second generation of ROS, which has a fundamentally better architecture that came out in 2017, right around the time... Both my co-founder and I decided then this is really a market opportunity.

Jan Becker:
We've seen so many companies... So, hundreds of companies use ROS and R&D but actually, none of those companies or researchers were able to move their ideas with ROS from a prototype to a product in a safety critical application. And that then became the business idea for starting Apex.AI in 2017, which is we've taken ROS, ROS is omnipresent in R&D, but we've now made it much more reliable. We've hardened it, we made it real time. We did that in throughout 2017, 18 and 19. And then, right when the pandemic hit in early 2020, we then actually started functional safety certification, which is a certifica... There are numerous different norms that describe what you need to do in order to develop functionally safe systems.

Jan Becker:
And functional safety is really defined as the avoidance of risk and the reduction of risk when it comes to safety critical failures in those applications. So, failures that could potentially lead to harm of persons or objects. And we then moved through that certification specifically for the automotive market in one year and now, since March 2021, so then over a year ago. We, both as a company and Apex.OS as a product, Apex.OS being the fork of ROS tool, which is now hardened for safety critical applications. We are both certified to ISO 26262 to ASIL D, which in practice means our customers, car makers, truck makers, but also, tractor makers or farming equipment makers can now take Apex.OS and build their product on top of it and such that it's then already certified, also for use on public roads.

Michaela Paukner:
Wow, your involvement with the autonomous and robotics community is so impressive. Just a couple of questions before we move on to talk about Apex.AI's involvement with AGCO. In the 1990s, when you were writing your PhD, were you thinking about autonomy for farm equipment at that time?

Jan Becker:
In the 1990s? No, not at all. Even though I always had a hunch that the introduction of such autonomous systems should actually be much easier in what we would, from an automotive perspective, call niche applications. So, applications where the volume is lower. We, in the world, make a 100 million passenger vehicles a year, about 10 million in the U.S., and also the largest car makers. Volkswagen and Toyota make about 10 million vehicles a year. One robust issue in those vehicles can have, obviously much, much more impact than if and when you move to applications with simply smaller volumes, such as trucks or farming.

Jan Becker:
In addition, fields or agricultural areas are simply easier to navigate. You typically have open space, you don't have tall buildings, blocked GPS reception, or lead to reflections of light and so on. So, the farming area is actually a great example of how much easier it is to build robust and reliable autonomous applications with the real customer value, customer in this case being the farmer who's able to work on fields much more efficiently. It's much easier to build applications in those areas than it is to go into the most difficult to master scenario right away, which is, for instance, driving in an urban environment.

Speaker 4:
Wanted to publicly recognize here for the first time, our 2019 most valuable dealership winner, Crystal Valley Cooperative. I would like to publicly present to you, for the first time, with the 2020 most valuable dealership award.

Speaker 5:
I'd like to publicly recognize, for the first time, the 2022 most valuable dealer, Jenner Precision.

Michaela Paukner:
You could be the next dealership. We announce as Precision Farming Dealer's 2023 most valuable dealership. Now, in its 11th year, our annual program recognizes the organization demonstrating the best in sales, service and support of precision farming technology. Dealers, manufacturers, and others are invited to participate by nominating top Precision Farming Dealers from across North America. Go to precisionfarmingdealer.com/mvd to nominate a dealer for our 2023 MVD award and help us recognize North America's premier precision farming operation. Now, let's get back to the conversation as Jan talks about Apex.AI's partnership with AGCO.

Michaela Paukner:
Apex.AI Is working with AGCO to add new capabilities to their autonomous saver farming robot. Could you tell us a little bit about the capabilities that will be added?

Jan Becker:
Sure. So, AGCO has selected our end to end operating system, which we call Apex.OS for their Xaver farming robot concept. What that provides to AGCO is, really, a base system, a base software onto which then AGCO can develop all kinds of different applications. So, the best comparison is actually, and one that I like to use, is the one to the smartphone. So, if you have an iPhone, you have iOS running on your device. If you have an Android phone, you have Android running on your device and then, Apple and Google respectively provide so-called software development kits, which is iOS SDK and Android SDK. And this SDK enables app developers, so those can be high school kids or those can be professional companies, develop applications running on that platform so that they all look the same. And so, that all the functionality common to all applications.

Jan Becker:
So, for instance, reading and sensor data from the camera, reading in data from the microphone, outputting Windows onto the screen. Many apps need a keyboard, so you have a virtual keyboard that always looks the same. So, all those common functionalities are abstracted in the software development kit. This is, in a nutshell, what Apex.OS is providing to developers of autonomous systems and solutions, in this case AGCO. So AGCO gets with Apex.OS, a robust, reliable and also already certified-based software for end to end operating system, which allows AGCO to develop their autonomous applications much, much quicker. And they don't have to worry about underlying software, they can entirely focus onto the application layer and bring autonomous applications currently being developed on their concept robot, Xaver, bringing those two market quicker and better and with more reliable results than before.

Michaela Paukner:
Okay. What are some of the potential applications that AGCO could develop on top of the Apex.OS-based software?

Jan Becker:
I can't speak on AGCO's behalf here, but some of the examples I can give you is that we have customers that have developed object detection with Apex.OS. So, using a lighter or using a camera to detect where objects are then doing collision-checking, so are there any objects, a robot in this case, an autonomous farming robot, could potentially collide with. Then, if that is the case, you would either stop the robot or plan a pass around it. And not just the application on top, what we also provide is a much, much more efficient and state-of-the-art framework for developers to develop autonomous applications. So, our customers, in this case AGCO, really have the advantage that they can get to market much quicker with autonomous applications. Because all the underlying framework, such as, again, iOS or... That's exactly what iOS and Android provide to app developers. They make it much, much easier than before to develop applications than without having such a robust and reliable framework. This is really the core value that it's provided to AGCO.

Michaela Paukner:
Can you provide an example of how another customer has used Apex.OS from start to finish and what our involvement looks like from the Apex.AI side?

Jan Becker:
So, I cannot talk about specific customers, but I can absolutely talk about application areas, customer areas, and what some of these customers do. So, we work with a wide range of passenger car, both OEMs and suppliers, with trucking companies, trucking technology developers, with shuttle and robo-taxi manufacturers. And some of the applications range from a fully autonomous level four stack, so level four meaning a vehicle can operate in a certain environment completely without a driver. We've also have customers that develop driver systems technologies, so level one, level two.

Jan Becker:
On top of Apex.OS, we have a company that develops a passenger monitoring and a passenger interaction system. On top of Apex.OS, we have customers evaluate and use Apex.OS for lane-keeping, for instance, which is also an application that is very close to farming applications, where you want the farming robot to do autonomous steering on the field so that the farmer can work or the operator of the device can work on other things. And then, focus on, for instance, turning around the vehicle or maintaining operations overall. So again, overall, all those applications essentially make the task, whatever the task is, either more efficient or more reliable or safer or the combination of all these three things.

Michaela Paukner:
It seems like a huge benefit to AGCO to be able to have all three of those benefits available to them right away. And how did Apex.AI first get involved with AGCO?

Jan Becker:
So that's actually a funny story. So, one of the engineering directors at AGCO Germany, he and I worked together in the same room during our PhD studies a little over 20 years ago. So, while I was working on autonomous vehicle, he was working on control of electrical machines. In the end, we did keep in touch over the years and at some point, he moved to AGCO in the U.S. And then, at some point again, transferred to AGCO in Germany, and he's now leading an R&D department there, actually the R&D department working on this autonomous systems. And then, there is another coworker who actually came from the same R&D lab where we two did our PhD thesis, and then also joined AGCO in Germany working on autonomous systems.

Jan Becker:
And then, they reached out at some point expressing interest in Apex.OS, after they heard of all the benefits that we provide. And also, an important aspect is what I've mentioned before. Almost everybody in the industry, not just limited to autonomous farming applications, but also to autonomous vehicle applications overall, is usually using ROS for prototype development already. And we then simply provide an easy and seamless and fast transition from a ROS, which is open source, from a ROS-based application, to a safe and certified Apex.OS application. Then also after we worked together for while, AGCO actually reached out and there is an investment opportunity. And so, AGCO then joined the Series B investment round, which we closed and then announced last December.

Michaela Paukner:
How has that strategic investment benefited Apex.AI?

Jan Becker:
So, it benefits us in the sense that by not just working with a customer, also having in this case, a customer invest in us really is a strong sign of the value that we provide as a company and the future that, in this case, the customer AGCO sees in us. So, it really provides us with validation and having, in this case, an agricultural customer invest in Apex.AI also provides us with industry validation that the applicability of our technology really goes just beyond cars. So, we had AGCO invest from the farming space. We, for instance, also had Airbus invest as an aerospace company, and Daimler Truck as a trucking company and Toyota invested before as a passenger vehicles company, and we've had a number of suppliers such as Continental, at [inaudible 00:25:09] also invest. And many of these suppliers also provide components and systems for commercial vehicles, such as farming equipment.

Michaela Paukner:
What would you say differentiates the Apex.OS base software for farm applications from other autonomous software for farming?

Jan Becker:
First and foremost, we are the only product on the market that is based on and compatible to ROS, which again is the system that most companies already use for prototyping. And since we are based on and compatible with ROS, we provide a very easily and seamless transition from prototype software and ROS to commercial production software using Apex.OS. That's differentiator number one, and then differentiator a number two is that we are also the only company on the market that provides, to date, fully-certified software to the highest level of automotive function safety, which is a ASIL D.

Michaela Paukner:
Okay. And then, what do you see as the future of autonomous farm equipment?

Jan Becker:
So, to be fair, I'm really an autonomous systems and robotics and software expert, and the specific farming application is not my strong suit. But what I would imagine is that there are so many just dirty and boring tasks in farming. So, driving a tractor on a field for 10 to 12 hours a day is just not a fun task, and once things are not fun and boring at the same time, people get distracted. You also see that in driving, once you drive long distances, people take up their smartphone, even though they shouldn't. And that is then when either reliability issues start to occur, so people get distracted, or also safety issues start to occur. Plus you enable, in this case, to operate as a farmer to maybe more efficient. So, for instance, a farmer can already do paperwork or be otherwise productive, managing logistics while being just driven back and forth on the field.

Jan Becker:
And then, for instance, once the tractor leaves the field, the farmer could take over again and then focus on driving the tractor back to the farm. So, that would be an example of a low hanging food where automation of just driving on the field is easy and can be achieved maybe in a shorter time, whereas than driving on a public road. Because then you also need public road driving approval. It's harder to achieve, so you would launch it as a product later on. So, in the meantime, there are certain tasks that the farmer still needs to fulfill, which is driving back on the road, but then for 95% of the time, where the tractor is driving on field, that could most likely be automated much sooner and easier launch to market.

Michaela Paukner:
Thanks to Jan Becker for today's conversation. If you'd like to listen to the episode featuring my interview with AGCO, Seth Crawford that I mentioned at the top of this episode, visit precisionfarmingdealer.com/podcasts, or check out our episode library, wherever you get your podcasts. Be sure to let me know what you thought about this episode by leaving a comment on the web story for this podcast or on Precision Farming Dealers, Facebook or Twitter. From all of us here at Precision Farming Dealer, I'm Michaela Paukner. Thanks for listening.