Starsky & Hutch API: Starsky Robotics Launches API to Dispatch Autonomous Trucks

Stefan Seltz-Axmacher
Starsky Robotics 10–4 Labs
5 min readAug 14, 2019

If you look up the definition of product in Merriam-Webster’s dictionary, this is what you will find: “Something (such as a service) that is marketed or sold as a commodity.” Well, today I’m excited to announce Starsky’s actual product: The Hutch API.

A little over two years ago in February of 2017, we came out of stealth as a developer of autonomous trucks. Yet we aren’t selling autonomous trucks. In June of this year we revealed that along the path to developing autonomy, we have built a trucking business. Yet Starsky isn’t really selling trucking — we’re not spending our time convincing shippers that we can solve their complex logistics problems.

Today I’m thrilled to announce Starsky’s Hutch API, which from our customers’ perspective is our actual product.

The Hutch API makes our trucking capacity accessible to freight brokers and major shippers without backoffice intervention, making the whole shipment process, well, autonomous. Rather than “selling trucking,” we allow partner brokers (and select shippers) to integrate directly with us, enabling them to bid for space in our trucks without ever needing to call a member of our team. To that end, a few weeks ago we integrated Hutch API with Loadsmart’s Automated Dispatch API to dispatch an autonomous truck to haul freight without anyone in either office talking to each other.

The Hutch API interface

“The Hutch API makes our trucking capacity accessible to freight brokers and major shippers without backoffice intervention.”

The systemic long-haul truck driver shortage has made many aspects of American trucking funky. It’s made U.S. trucking incredibly fragmented (with 1.3m fleets operating 3.5m trucks, where the 90th percentile trucking company has fewer than six trucks). Large shippers can’t realistically work with so many companies which is why they partner with 18,000 freight brokers to aggregate capacity. The freight brokerage industry is itself an interesting space to innovate and has startups like Loadsmart, Convoy and Transfix working to out-innovate industry titans like CH Robinson. Just like how the industry has more demand than supply, traditional trucking companies like Schneider almost always have more sales than they have capacity and so they too operate brokerages (like Schneider Logistics).

Like all marketplaces, trucking brokerage is a two-sided problem, but one with ever-present demand on one side and ever-scarce capacity (long-haul trucks) on the other.

Starsky was founded to solve the second half of that problem — we’re creating capacity in the capacity-constrained trucking market by making trucks drive without a person onboard. Rather than trying to sell those safety-critical robots to traditional trucking companies or to enter into complex contracts with shippers, we’re making that capacity available to the market via the Hutch API.

This is a big deal for a number of reasons. A typical trucking company might have 5 individuals that spend one cumulative man-hour to dispatch each truck for each load every day. If Uber were run like that, they would need ~2.5m dispatchers as opposed to their 22,000 employees in total. Trucking dispatchers aren’t negotiating rates; they’re just deciding which trucks should haul which loads and relaying that information to the drivers. As a tech company, this sort of inefficiency is offensive: if we operated 100,000 autonomous trucks this way, we’d need 31,000 dispatchers. We would have solved an incredibly hard problem while creating a silly one. The Hutch API solves that. It also frees our team to focus its time on helping drivers and dealing with the complex / non-routine issues that come up in trucking every day.

“If Uber were run like a trucking company they would need ~2.5m dispatchers.”

Needless to say, releasing the Hutch API also helps us to scale. We’ve built a rockstar 6-person trucking operations (TruckOps) team for our 40+ truck trucking business, but I don’t want to triple the size of that team if we triple the size of our fleet by year’s end.

Finally, the Hutch API allows us to automate much of the manual work that comes with testing an autonomous truck. The logistical complexity of keeping our regular trucks moving has effectively painted a picture of just how difficult it will be to do so with the autonomous fleet. Testing a few autonomous trucks is uniquely different — we do things like repeatedly drive a single portion of a route to characterize connectivity at different times of day or string together a series of routes to measure their relative difficulty. Doing that with one field test truck is hard and often involves coordinating 4+ people on the team. As we scale the number autonomous trucks in the field, that complexity will increase magnitudinally. The Hutch API helps us more effectively connect the tasks from a member of the engineering team to a growing number of trucks in the field without as many redundant conversations and without requiring us to expand our testing team proportionally to the autonomous fleet.

Endavr, Starsky Robotics’ truck that completed a 9.4mi unmanned test on a public highway in June 2019

One of Starsky’s core values is “autonomous products, processes, and people.” While the meaning behind autonomous products and autonomous people should be self-explanatory, what’s often unclear is that we want to make all of our processes as efficient as possible. We don’t just want to build autonomous trucks: we want to build autonomous trucking. And a lot of the latter has to do with automating workflows like accepting lanes from brokers not via phone calls, but instead via automated dispatch APIs like Loadsmart’s. When we have those 100,000 trucks we want to be able to add 5,000 more without adding a single operations person. We’re one step closer to that with the Hutch API.

“We don’t just want to build autonomous trucks: we want to build autonomous trucking.”

The Hutch API will increasingly be how we interact with the market. Integrating with Loadsmart is our first step towards moving our day-to-day operations from phone calls between people to API calls between servers. We are actively looking for additional partners to get loads from, and new announcements will not be long in coming.

Keepin’ on Truckin’

-Stefan

--

--