Autonomy’s Sensible Older Sister (Why we invested in Starsky Robotics)
In the waning days of 2015, I opened my inbox to see a note from someone named Stefan Seltz-Axmacher. He had been reading FoT and wanted to tell me about his new project: ‘After 2+ years evaluating startups, I’ve started a driverless trucks startup.’
I remember reading Stefan’s note, shutting my laptop and immediately walking into a lunch meeting with an experienced investor who went on and on about how he didn’t respond to inbound emails from entrepreneurs unless they come from warm introductions. He seemed pleased with himself.
What a horse’s ass, I thought.
If you want to trade on the future, you can’t discriminate based on the past.
It was soon thereafter when Stefan and I met in person. I learned as much about commercial trucking in that first hour than I had in the previous 12 months. In fact, I was rather dumbfounded at some of the minutia he had uncovered and how he was already thinking about how to improve on parts of it through his startup. We promised to stay in touch and it was some time later that I met his co-founder Kartik Tiwari.
The impactful thing about their point of view early on is that they fundamentally wanted to build autonomous trucks from within the trucking industry. This was not set out to be a Silicon Valley version of repeal and replace. They had an appreciation and understanding of the unique way that trucking companies operate today.
Trucking is Already A Microservice Transportation Platform
Most people don’t realize that trucking today is already unbundled into a dozen or so microservices. Trucks that you pass on the highway are a rolling assemblage of service plans: you can buy a truck for a few dozen cents per mile, you can buy a maintenance plan for about ten cents per mile, insurance is about five cents per mile, tires are about three cents per mile. Fuel and drivers remain the largest expense, at roughly 50 cents per mile and 40 cents per mile respectively. As it turns out these aren’t simply the costs of running a truck as measured by a small unit, this is actually how many trucking companies structure their operations. They buy services per mile from manufacturers like Volvo Trucks and service providers like Penske. If you want to start a trucking company tomorrow, you could bundle up many of these services and deploy a fleet by next Tuesday.
Most people also don’t realize that the driver shortage in commercial trucking — estimated to be in the red by roughly 100,000 drivers — has some unique forces at play. Drivers with seniority tend to prefer shorter routes near home (sleeping at home is a great benefit), while newer drivers tend to get tasked with long-haul routes (although not too new — interstate routes have a 21+ requirement). That is a tough thing to do to new drivers who still might be figuring out if they want to remain in the industry. Stefan’s view was to lean into this market dynamic and actually come up with ways for the best drivers to remain close to home.
Fast forward to the Spring / Summer of 2016, we at Trucks had made our first investment in Starsky (we are the company’s largest investor, although as of today Stefan and Kartik have financed the company with nearly $4M from others such as Data Collective, Sam Altman and Dynamo) and they had been accepted into the summer class for YCombinator.
Last week Starsky emerged to the public for the first time. The cover was ripped off to reveal an autonomous vehicle technology that — perhaps paradoxically — highly leverages humans for complex tasks.
Isn’t the point of automation to eliminate humans? I would submit that the point is to create the safest possible solutions within the shortest amount of time. With that in mind, the way Starsky is using tele-operation will become more important in the autonomy discussion over the coming years. (see Alex Roy’s post about Starsky for The Drive here)
In the U.S., there is something known as the ‘hours of service’ regulation, which states that truck drivers can’t drive more than 11 hours in a day and be on service for more than 14 hours with a day. It’s a smart thing to maintain safety, but it also means that trucks sit fallow for a significant part of the day. A clever way around this regulation used to be to partner with another driver and let him or her rest in the sleeper berth while you drive. But driver shortages have made that less common.
Starsky’s solution is resourceful: they use automation during highway tasks (although speeds are higher, these tend to be less complex in terms of pedestrian and bicycle avoidance, plus the roadways don’t enter intersections or encounter cross traffic), but fall back to a human in complex situations such as the first and last mile when a truck is entering a low-speed environment like a warehouse truck yard.
Tele-operation is an elegant MVP for automation today
The way they accomplish this is by tele-operation, or by allowing a human to operate the vehicle in a remote capacity. A contemporary analogy would be the military’s UASOs, or unmanned aerial systems operators, aka drone pilots. These individuals sit at a mirrored terminal in a remote location and ‘pilot’ the vehicles as if they were sitting in them.
Using tele-operation in trucking automation is interesting for two reasons: 1) it allows the vehicles to operate for 24 hours straight (by changing out remote drivers, hours of service limitations do not apply), thus making the truck itself vastly more economical and safer, and 2) it allows for the industry to retain great truck drivers by letting them drive each day and sleep at home at night.
It’s quite possible tele-operation will become a more common backup for some automated driving systems. I describe it as autonomy’s sensible older sister. This ‘human substitute’ allows today’s systems a reasonable fallback should they encounter something too complex for their current software and hardware paradigms.
Starsky was formed to build a new type of trucking business model. I’m happy we’re backing them as they grow.