An Autonomous Trucking Startup Stays the (Very Long) Course

(Bloomberg) -- Nancy Sun, 37, is chief engineer at Ike, the autonomous-trucking startup named for President Eisenhower and his interstate highway system. Sun and her two co-founders, Jur van den Berg and Alden Woodrow, started the company in 2018 after leaving Uber Technologies Inc.’s autonomous-vehicle unit. An MIT-trained engineer, Sun has been working on self-driving trucks since she joined Anthony Levandowski’s Otto in 2016. Before that, she worked in the special projects group at Apple Inc. and before that for the defunct electric-motorcycle startup Mission Motors.

Trucking has become an area of increased interest and activity in the autonomous-vehicle industry in the last few years as efforts to build self-driving cars have proved more costly and time-consuming than expected. Ike competes with other startups, including TuSimple, Kodiak Robotics, and Embark Trucks, as well as industry pioneer Waymo, which increasingly views trucking as its fastest path to market. Yet the future of self-driving trucks is also uncertain. Last month, trucking startup Starsky Robotics shut down after four years in operation. In a Medium post announcing the company’s end, founder Stefan Seltz-Axmacher wrote that “supervised machine learning doesn’t live up to the hype.” The economic fallout from Covid-19 adds to the challenges facing the industry, but also, potentially, to the demand for autonomous technology.

Last year, Ike announced that it had raised $52 million in a funding round led by Bain Capital Ventures. The San Francisco-based company began with software licensed from autonomous-delivery startup Nuro. The two companies now operate independently, with Nuro working on last-mile delivery and Ike on long-haul trucking. Ike’s plan is to build technology that can drive on the highway, with human drivers taking over the shorter trips to and from depots. By Ike’s own projections, autonomous trucks running day and night without breaks for sleep would add roughly as many jobs for short-haul drivers as they take away from long-haulers, while keeping drivers closer to home and making roads safer. That vision, Sun says, remains years away. Ike is taking a long-term approach. It does not test on public roads or carry commercial cargo. Instead, its team of 70 is focusing on building an autonomous system that truck manufacturers, freight companies, and independent owner-operators will trust.

We spoke with Sun in mid-March, just after California Governor Gavin Newsom ordered the state’s citizens and nonessential workers to stay at home.

First, how are you?
We are doing well, all things considered. We were pretty proactive with our team about transitioning as many people as possible early on into working from home. And so the shelter-in-place ordinance actually didn’t affect us too much.

What can Ike keep doing during shelter-in-place requirements, and what has to stop?
Actually, most things can keep going. The investments that we’ve made in simulation and software infrastructure enable us to test really quickly with data that we’ve collected, instead of having to be in the real world doing real-time testing. The bank of miles that we have is still real-world stuff, so we aren’t even really pausing real-world stuff. It’s just pausing being able to increase that bank of data. That’s where simulation helps us.

We think of testing in two buckets. The first is manually driven miles on public roads. The second is all the testing that we do at private test tracks. We are not operating our vehicles in automated mode on any public roads. We have oriented the engineering team around trying to use simulation and use the test track as our primary areas of automated testing, and then using real-world manual miles as data to feed both of those things.

How long you can go without new real test-track or manually driven miles?
At some point we will need more data, but for the foreseeable future we’re totally fine.

What did you learn from your time at Apple and Otto and Uber ATG?
There’s this broader narrative around automation and the robot apocalypse, but we really have an approach that automation can improve everybody’s lives as long as we do it thoughtfully and responsibly. So from the start at Ike, we’ve always been asking ourselves how can we help truckers, not replace them. It’s something we think everybody in the industry should be asking. What we’ve found is that some folks are more interested in automation either for the sake of disruption or the market opportunity within trucking. We’ve spent a lot of time with truckers. What we’ve heard from them is that the highway is the hardest part of the journey.

We believe we can help create a future where the truck driving job can look very similar to how it does today but be safer, be more productive, and then also be higher value. We also think about the technology in the space as a complete system and not just a software problem. I think there’s a trend throughout the automation industry to say, “If we just solve the software problem, then everything will be fine.” There are different pieces that go into developing an automated-truck product. Software is one, but there’s also a hardware component. There’s a regulatory piece to it. There’s a broader umbrella around safety. We’re focused on being able to build all of the pieces, not just one.

Are there going to be wide-scale autonomous, long-haul trucks on highways? And, if so, when?
I think so. It’s still years away. Many people have talked about deadlines, and those deadlines have already come and gone. And so one of the things that we’ve had to be realistic about is just resetting expectations on the timeline. Building an automated-trucking system is really difficult, and there’s so much work needed to build something real. The temptation—and I’ve done this in my past, at various steps along the way—is to kind of hack something together, to get it up and running really quickly. But we also know, because we’ve done that in the past, how important it is to be able to measure progress towards a real product. And that’s not in the number of real-world miles that you’re driving. It’s not in the number of vehicles in your fleet. It’s not measuring by the number of loads you’ve hauled for revenue. We are going to talk a little bit more about our metrics in the near future. I’m reluctant to give a preview, but we do want to share some of our thoughts around metrics and what are the right things to measure.

Why did you choose to work on trucking? What are the advantages over other types of autonomous vehicles?
The trucking industry faces huge challenges, but it’s also one of the most important industries in our economy because literally everything moves on a truck. There’s a shortage of drivers. There are all these new regulations. The accident rates are increasing. And that’s all coupled with a rise in e-commerce. Truck drivers end up often being overworked and are actually doing a fairly dangerous job. We think that automated trucks can really help address many of these challenges. You can imagine that a truck without a driver can operate double the hours every day. It can reduce the thousands of lives that are lost in preventable accidents. And then our highway focus means that drivers still have a really valuable role to play in moving goods to and from the highway. When it comes to the application of self-driving technology, there are advantages to focusing on highway operation instead of passenger cars in cities—there are just fewer variables at play. Think of all the decisions you have to make and the unpredictable hazards that are involved when driving in a city.

What will your end product be?
This is another area where we have a fairly unique approach. Most folks working in the automated-trucking industry are focused on being a fleet or outfitting their own fleet of vehicles for operation. And our approach is to partner with the existing industry—the Tier 1s and [original equipment manufacturers] and fleets and shippers. We want to work with them to build a product that they think is actually going be useful.

So your customer would be a fleet operator or maybe a Tier 1 manufacturer?
We’re pursuing many different strategies on that path. Without being able to get into the details of the model right now, we’re focused on the partnership piece rather than zoning in on any one specific area.

Reducing or eliminating monotonous highway miles sounds great for truck drivers, but where is value for your customers?
There’s actually an opportunity to create more jobs doing short-haul because of the increased efficiency that comes with automation when applied to long-haul. The other aspect is around being able to increase the utilization of a really expensive fixed asset [trucks]. Today, truck drivers are limited in their hours of operation—well, perhaps not in the time of coronavirus, but in general.

Is the plan to be able to operate in all conditions?
The idea in the very long term is definitely to be able to deploy technology nationwide. In the short term, in order to go to market, we will want to constrain the operating domain.

What are some of the things that you’re ruling out for now?
The hardest things to do are going to be mountainous terrain when the weather is terrible. Those are areas that we’re going to want to avoid. I think there’s a pretty good match for launch in the geographies in the Southwest just because of the combination of good weather and high freight volumes.

What’s your take on Starsky shutting down?
Doing startup is hard. And developing technology in this space is hard. I think we will see more of the industry start to shake out and consolidate over time. There are hundreds of lidar companies right now. You can’t even count on your hands and toes the number of automated-vehicle companies that are out there. So I think there’s likely more [consolidation] to come in the future.

©2020 Bloomberg L.P.