Commentary: Real trucking factors in autonomous vehicle development

Starsky autonomous truck

The views expressed here are solely those of the author and do not necessarily represent the views of FreightWaves or its affiliates. 

I was fortunate to be afforded the opportunity to be an operational leader in an autonomous truck development company by the name of Starsky Robotics. Starsky was a Silicon Valley-based start-up with an extremely talented engineering team and clear path for technology advancement. Being a career trucking guy with an inside track to emerging technology, many people were curious as to what we were actually trying to accomplish. I was frequently asked by trucking colleagues, “How well does it really work and how soon will self-driving trucks take over the trucking industry?” In fact, it’s not only a question asked by trucking executives, it’s a very hotly debated question in both technology and logistics circles alike.

Recently, the founder of Starsky Robotics, Stefan Seltz Axmacher, sparked a huge on-line debate with his blog on I purposely chose to refrain from comment in that debate, because it was more directed toward technology and technology limitations. I’m far from being an expert on whether a company should use radar, Lidar, HD cameras, machine learning or even artificial intelligence. My expertise is in trucking, and after spending over 30 years leading operational teams to maximize the efficiency of logistics, my earliest question was a simple one – “If you’re developing self-driving trucks, why would you need a trucking company and why would you need someone like me to run it?” 

It didn’t take long for me to understand the answer to this question, as we developed and refined our objectives as a company… keep it simple by limiting the variables, don’t try to be everything to everyone, improve the work-life of the driver, achieve safety through extensive validation utilizing professional drivers, operate a trucking-based business plan and lean on trusted relationships in the existing trucking industry to further the cause. It was these basic concepts that allowed me to fully commit to what we were attempting to do at Starsky. I truly believe it’s these basic ideals, not the technological debate, that provide the real path to operating self-driving trucks, on some scale, in the not too distant future.  

A Starsky autonomous truck drives down the highway at dusk.
(Photo credit: Starsky)

Let’s explore each of these things in a little more detail, starting with the most hotly debated technology topic.

1.      Keep it simple by limiting variables

Have you ever heard the question, “How do you eat an elephant?” Well, the autonomous vehicle (AV) development industry is truly an elephant. There is a huge amount of competition but there are also massive differences in approach – everything from should we build a truck from the ground up, should we retrofit an existing truck, should we utilize current original equipment manufacturer safety features, where should we operate the trucks, what truck manufacturer is best for AV, should we be focused on electric vehicles as well, do we focus on driving in bad weather, what about tricky roads, mountains, hills and heavy traffic? At Starsky, we employed a very simple concept – we decided to eat this elephant one bite at a time. Prove that you can safely handle difficult, but achievable tasks first; you can always expand functionality once you prove that you can safely operate in good weather, with proven equipment, on the same route, with an experienced driver monitoring progress along the way. 

It was this basic approach that allowed us to safely operate a loaded truck, without a driver in the cab, for almost 10 miles in Florida last Father’s Day. This first bite at the elephant was something that no other AV company had accomplished. This achievement opened the door for us to expand the route, expand features and basically take the next bite at the elephant. Bottom line, by reducing the scope, rather than trying to deal with every inevitable scenario, we made real progress toward AV deployment, simply by limiting variables. Our approach also enabled me to sleep comfortably at night, secure in the confidence that we could operate on the same stretch of road every day with less than a one in a million chance of an accident – an amazing accomplishment that’s only duplicated in the very best of professional drivers.   

2.      Keep the safe, professional driver in the equation, but ultimately not in the cab

I think that everyone would agree that every AV company requires the input of professional drivers to be able to solve the self-driving dilemma. Leaving them out or thinking that they can just be replaced in a couple of years is not a recipe for success. I was once told by a reporter that Starsky’s articles elicited the fewest number of angry responses from drivers. Why, you may ask. 

Starsky was not focused on eliminating the driver. We were focused on engaging with professional drivers to develop technology that would improve the life of the driver. Autonomous development companies that ignore the continued role of the driver, do it at their own peril. The driver is a critical element in the development of this technology and thinking that we will eliminate all driver jobs with autonomous trucks is similar to thinking that we can eat the elephant in a single bite. In my opinion, we will never truly replace the truck driver. AV companies that operate on the idea that drivers will eventually not be needed in our lifetimes are kidding themselves. Those companies will fight an uphill battle on both the technology and general acceptance. The industry will not embrace technology that promises to eliminate all trucking jobs entirely and technology will not progress enough without critical input from valuable professional drivers. 

Starsky was successful in building an efficient 50-truck trucking company and we gained acceptance in the industry via our committed driving force. Drivers were proud to further the concepts because we had a place for them in the future. Our drivers were our best advocates, helping us to gain fans with customers, vendors and even other drivers. You just cannot dismiss the momentum you can gain through your drivers with vital feedback and committed engagement for a better future. The industry’s driver shortage is a real problem, but not because there are not enough people needing a decent paying job. The real reason lies in the fact that truck driving is not the job of choice. We sought to change the job to make it a job of choice.

A truck without a driver moves down the road.
(Photo credit: Starsky)

3.      Achieve safety through driver feedback and controlled repetitive validation

This topic solicited significant amounts of debate when Stefan published his blog a couple of weeks ago. Many were upset with the assertion that machine learning was not going to rapidly solve all of the challenges associated with operating a truck without a driver. I had a slightly different take on it as a trucking veteran, and I think it aligns with many of my colleagues in the trucking industry. Rapid deployment of AVs will not happen on every road, in every condition, for quite a long time but when an AV operates on a single lane over and over again, day in and day out, certain things can be clearly defined that significantly increase the potential of safely and consistently operating that same lane without a driver in the cab. 

Simply stated, technology is not required to be completely relied upon to determine what obstacle is in the road, but rather to determine if an obstacle exists. Safety measures can then be employed immediately to react in an appropriate manner. Shifting focus to earlier object detection and avoidance, rather than trying to determine if something is a tire or a trash bag, is a safety concept that most anyone can understand. Reaction times improve significantly in the absence of driver fatigue, driver distractions and improved visibility. Seconds count when it comes down to safety and reaction time is critical to having safer options. It only takes a single misclassification of an object by a driver or by a computer for a vehicle to be in an accident, and that’s just unacceptable. I’ve long said that the future of AVs relies on providing a safer alternative to the very safest driver on the road and that is a tall order. 

There’s no argument that technology has and will improve safety on our roadways, but severe accidents, public perception and general fear has the real potential to sideline any autonomous vehicle company in an instant. It’s my opinion that autonomous companies that grasp the concept of introducing self-driving trucks in very controlled environments will achieve their outcomes sooner than those that try to operate in every environment, on every road, at every time of the day. Too much reliance on developing perfect technological intelligence will just slow the progress toward implementation to a crawl.

4.      Develop and execute a reasonable trucking-focused business model

There is a general belief among some in the AV industry that long-haul lanes will be the best lanes to deploy AVs. I’ve even seen pitch decks that claim huge profit opportunities by operating AVs exclusively, 24 hours a day in long-haul lanes Southeast to Southwest or from the Midwest to the West Coast. This just isn’t realistic at this time in the trucking industry. 

At Starsky, we had a business model that would allow us to eventually make money hauling mainly short-haul and regional freight. We focused on our costs, productivity and rates to find balance. Our regionalized/short-haul approach had positive implications for a number of reasons. First of all, having direct access to each autonomous truck on a daily basis is important and it’s also achievable in a short-haul or regionalized configuration. Things like daily inspections, preventative maintenance and fueling can all be accomplished easily when the truck can be seen daily. These things will be very costly problems for long-haul AV companies to solve.  

Secondly, the freight rates on long-haul lanes are much lower because of competition with railroads. Some dry van rates can be so low that they don’t even cover the fixed and variable costs of running a truck even when you don’t include driver pay. Alternatively, short-haul and regional freight provides rate opportunities that allow the individual unit economics to make sense. AV companies need to come to grips with the fact that trucking companies operate based on cost per mile, fixed and variable cost structures. Trying to compete with railroads on costs or thinking that customers will be willing to pay you more with an autonomous truck just doesn’t compute.  AV companies that ignore the basic unit economics of running a single truck will be slow to grow and will not provide a real business case for adoption, hindering rapid deployment of AVs.

A sign states "LIDAR test range."
(Photo credit: Jim Allen/FreightWaves)

5.      Above all else, develop a win/win value proposition

Despite achieving significant technological milestones, Starsky was unable to secure additional funding in November 2019. I’m of the strong belief that it had nothing to do with technology; instead it was about the immediate value proposition for investors. I can’t help but surmise, if Starsky would have more aggressively pursued investment partners that would eventually benefit operationally from AV development, things may have been different. 

Billions of dollars have already been invested in AVs by venture capital firms, and the pot of gold at the end of the rainbow still eludes them. Venture capital investors expect a win proposition and they’re only stakeholders in relation to the potential of receiving a significant cash return on their initial investment. I believe the real opportunity for AV development companies is to seek out and develop relationships with investors, companies and individuals that will truly benefit from the end product. In trucking, those customers and stakeholders are drivers, vendors, shippers, consignees, brokers, trucking companies, government agencies, recruiting companies, and training companies, just to name a few. Developing win/win partnerships with these customers and stakeholders will not only provide opportunities for tangible cash investments, they also provide valuable collaboration on safety, increased operational viability, valuable marketing, cost reductions, critical technology input and overall credibility in the industry.

In conclusion, I’ve had the opportunity to reflect on what it’s going to take to bring AVs in scale to a highway near you. We all need to face reality – autonomous trucks are inevitable – in fact, they’re already here. I personally cannot predict with a strong degree of accuracy when they will notably revolutionize the trucking industry. What I can predict is – without the direct involvement of the trucking industry as a whole, the timeline will most certainly be longer. I’ve seen with my own eyes how well AV technology can and will work, and I’ve concluded that it’s not all about the best technology, the perfect approach or even the massive amounts of validated autonomous code stored on servers. 

It’s still about an industry with real people, mutually beneficial business relationships, conservative cost per mile financial plans, and significant real-world challenges that won’t immediately be solved with the latest technological advancement. I’m 100% on-board with deploying safe, autonomous technology that improves efficiencies in the industry but until that day comes, we must collaborate closely with an industry wide expertise that’s managed to deliver over 70% of our products for nearly a century.