The Hub-to-hub Model May Drive Adoption of Driverless Long-haul Trucking Sooner Than Other Driverless Use Cases
on Aug 16, 2022
Projections for the rollout of driverless cars have dramatically shifted in recent years. Around 2015, the industry was predicting that autonomous cars would be available "before 2020." Now they are saying "after 2030." What does that mean for when driverless trucks may become available? We explore some reasons that we may see driverless long-haul trucking, in a hub-to-hub model, well before driverless cars are plying our city streets.
Full Article Below -
Optimistic Predictions for Self-Driving Vehicles Rollouts Have Failed to Materialize
Balsamiq, the popular user-interface wireframing tool, renders prototypes that look like a rough sketch. That low fidelity look is on purpose. One reason (among many) is if executives see something more polished that looks like a final product, they can easily assume it’s 90% done and ask why we can’t ship it now. During the 2010s, autonomous vehicles seemed to suffer from an extreme version of this “90% done” phenomena, where it looked like they were practically ready for widespread use, when in reality the longest and hardest work remained ahead. It seems that the majority of the auto industry fell prey to the ‘90% done’ fallacy for autonomous cars around mid-decade (c. 2014 to 2017) when most major carmakers were predicting1 they would have fully autonomous vehicles by 2021 or sooner … in some cases much sooner.
Here we are in 2022 and you still cannot buy a self-driving car anywhere.2 According to MotorTrend, “No vehicles available for sale in the U.S. today are self-driving. We may see Level 3 self-driving cars in the next five years, but they would only work in very limited conditions. There are still many problems to tackle before Level 4 or 5 autonomous cars can become a real thing.” Forbes states, “there’s no realistic chance that full-on self-driving will be available before 2030, and then only in a tiny number of top-of-the-range sedans and SUVs.”
A Pathway to Autonomous Trucking
So, what do these sobering predictions mean for autonomous trucks? For non-highway, public-road trucking, the timeframe for autonomous vehicles (AV) trucking is probably not much different than for AV passenger cars. However, it is quite a different story for the use of autonomous trucks on private property and hub-to-hub long-haul trucking. We’ve been predicting for a while that these will be two of the first areas where autonomous vehicles take hold.
Autonomous Mining Trucks Leading the Way
In fact, autonomous mining trucks have already led the way, having been in commercial operation for almost 15 years. Back in 2014, we wrote that Rio Tinto had been using autonomous, driverless mining trucks since 2008, at the time having already driven almost 3 million miles, and hauled nearly 250 million tons of ore. Since then, the use of driverless haul trucks for mining has exploded. Caterpillar alone claims that their autonomous trucks have driven over 90 million miles and hauled over four billion tons of ore.3 There are now over 1,000 autonomous mining trucks worldwide.4 While that accounts for ‘only’ about 2% of all mining trucks, the sales of driverless mining trucks is increasing at about 40% per year. That torrid pace of adoption is being driven by 20%-30% improvements in productivity, 20% reductions in operating costs, and improved safety (autonomous mining trucks don’t get tired or lose focus like human drivers do).5
There are several reasons that adoption of autonomous mining trucks is way ahead of other types of driverless vehicles. For one, these trucks operate entirely on private property, so are not subject to the regulation or challenges of vehicles on public roads. Also, mining trucks drive highly repetitive routes, requiring less of the ad hoc decision-making which can be difficult for driverless vehicles (more on that below). Mining companies have more control over the conditions these vehicles face, such as providing clear standardized markers and/or electronic locating systems for navigation and enforcing policies for the behavior of other vehicles and pedestrians on the site.
Use of Driverless Trucks in Ports on the Rise
Similarly, driverless trucks are already being used in ports, primarily in China. HERE Technologies predicts there will be 6,000 to 7,000 level-4 autonomous trucks in operation in Chinese ports by 2025. The motivating benefits are similar to those in mining (increased productivity, reduced operating costs), as are the reasons autonomous vehicles work well—highly repetitive routes in a small, well-defined, privately-owned piece of land with a high degree of control over the behavior of onsite personnel and equipment—contrast that with the relative lack of control over the behavior of people and vehicles on public roads.
Hub-to-hub Long-haul Trucking Will Lead the Next Phase of Autonomous Vehicle Adoption
Beyond the private-property uses cases described above, we predict that hub-to-hub long haul trucking will be one of the first other use cases where driverless vehicles are widely adopted. In this model, a human driver picks up a load at its origin site (e.g., a warehouse or factory) and drives it to a nearby hub (aka ‘transfer point’), typically on an interstate highway on the outskirts of town. At the hub, the driver unhitches the trailer from his tractor unit and hitches up the trailer to a driverless tractor unit. Alternatively, the driverless tractor hitches itself to the trailer as described in Truck Talk: Autonomous hitching edition. The driverless tractor then drives the load for the long-haul ‘middle-mile’ leg of the journey, from the origin hub to the destination hub, where this same process happens in reverse, to be delivered by a human driver the last mile to its destination.
Figure 1 - Outrider's Autonomous Hitching Process (Source: FreightWaves)
Back at the origin hub, once the driver has unhitched their load, he/she can drive their own tractor unit to the inbound area of the hub, where inbound loads are received. They hitch up their assigned inbound load (which an autonomous tractor has dropped off) and drive it to its local destination (e.g., a local distribution center or retail store or business) for delivery and unloading.
Hubs will have mechanisms for refueling (or recharging/battery swap)6 of the long-haul autonomous trucks while the vehicle is between loads—i.e., after the vehicle has dropped off the inbound load, but before the vehicle leaves with the new outbound load. The equipment and processes at the truck stops/fueling stations on the interstate will need to be enhanced and personnel trained to provide refueling to autonomous trucks for routes that exceed the range of the truck.
Removing the Most Stressful Link of the Trip for Truckers
The hub-to-hub model of autonomous trucking takes the most lifestyle-unfriendly portion of the end-to-end trip and lets machines do it instead of human drivers. Long-haul truck drivers are away from home for days (or even weeks) at time, driving endless miles, often alone, with the unhealthy lifestyle of sitting for most of the day. This lifestyle may suit some, but for most, the loneliness and isolation from family, friends, and city life is stressful and potentially depressing. The lifestyle challenges of long-haul trucking are one of the reasons there is a chronic shortage of truck drivers.
Interstate Highways Are Much Easier for Autonomous Vehicles to Deal with Than City Streets
There are a number of reasons that interstate long-haul trucking is easier for autonomous vehicles compare to off-highway/in-city driving. Interstate highways tend to consist of long, relatively straight stretches of roadway with no intersections and relatively infrequent on- and off-ramps (compared with the shorter distance between intersections in a city). They tend to have well-marked lanes and fewer variations in conditions compared to off-highway driving. Those characteristics make interstate highways considerably easier for autonomous vehicles to operate than city streets.
AI Struggles to Anticipate Human Behavior (Drivers, Pedestrians, Cyclists, Children)
Autonomous vehicles work well in repetitive situations, where they can learn the routes, and face relatively few new situations and conditions to deal with. In particular, driverless vehicles have difficulty interacting with human drivers, pedestrians, and cyclists. Interacting with humans is especially common and difficult in cities, where an autonomous vehicle may often be called on to read the subtle signals of the intentions and probable behavior of humans. One example is when approaching and crossing an intersection, especially with four-way stop signs, or the most challenging of all, the unprotected left turn.7 In these situations, human drivers are looking at the faces of oncoming drivers and pedestrians, to try and determine whether it is safe to make the turn or cross the intersection. Furthermore, drivers often communicate explicitly with each other, for example by waving the other driver through the intersection. Even for human drivers, the left turn is difficult, one of the highest scenarios for accidents. All of these scenarios are very difficult for AI to handle or emulate.
Other examples include discriminating between a vehicle that is trying to parallel park vs. one that is double-parked and unloading. Or determining the intent of pedestrians on a corner or standing by the side of the road. Figuring out if that person is about to step into or cross the road or is just going to stand there as you pass by requires complex reading of their body language and facial expressions, and what that person is paying attention to—all things that humans can do reasonably well, but machines find extremely difficult. On an interstate highway, there are far fewer instances of this kind of ad hoc decision making needed (perhaps when a vehicle is broken down by the side of the road). In a city, this type of complex decision-making happens all the time. That is a major reason that driverless cars are not ready for prime time, but widespread adoption of autonomous driving on interstate highways could be here sooner.
Regulation and Insurance Likely to Accommodate Hub-to-hub Routes Sooner Than Unrestricted City Driving
Regulation and insurance are often (correctly) cited as obstacles to AV (autonomous vehicle) adoption. We expect both of these institutions to accommodate the rollout of hub-to-hub trucking within a reasonable timeframe. Regulators have already approved AV truck pilots for a fixed set of routes in various states. One could envision regulators approving specific routes for specific vehicle fleets, once those fleets have proven to be consistently safe and reliable on that route. Perhaps regulators might define a standard number of incident-free miles (or incident free round trips) after which fully driverless mode gets approved for that vehicle/route combination. This per-route/fleet approval will eventually give way to broader approvals, as fleets prove their ability to reliably and safely handle a broader set of conditions and routes.
Insurers will need new models for determining liability and seeking compensation. Truck routes that cross state lines will require vehicle and insurance regulations to be solved and coordinated across the end-to-end route. Driverless trucks would need to be legal and insured in all states crossed by the end-to-end route. We may see entire regions that become AV truck-friendly sooner, in particular in the Southwest.
Some types of sensors get confused when faced with heavy rain or snow. Snow often obscures lane markings. Vehicles need to adjust their driving to account for road surface conditions, such as wet, snow-covered, or ice-covered roads. They need the ability to detect black ice and deal with high winds. Road flooding can require judgement calls, where even human beings can fail. One reason that we see so much of the early testing and production of AV truck routes being done in the Southwest is that weather conditions are much more favorable there. Ultimately, autonomous truck manufacturers have to solve weather-related driving challenges. Autonomous trucks that can only be used when the weather is good are not scalable, even in the Southwest. You can’t expect fleet owners to cease operation or suddenly find a bunch of human drivers whenever it rains, even if it happens somewhat infrequently.
There have been some seemingly promising developments in autonomous winter driving. According to this recent article, “Embark Trucks, an autonomous vehicle company powering self-driving semi-trucks, successfully completed initial winter conditions tests.” Their CEO stated, “Embark has a unique alternative to maps that we call 'vision map fusion, and it worked very well in the vast majority of the winter conditions that you would expect in the northern United States." It should be noted, the article did not provide any details of the extent or conditions of these tests, nor a measure of the performance of the vehicles.
AV trucks require extra-meticulous inspection, maintenance procedures, and increased predictive maintenance, to minimize breakdowns, since in-trip mechanical failures are harder to deal with when there is no driver. In addition, remote monitoring is critical to detect and alert when a vehicle has broken down. The truck needs to be capable of knowing how and where to pull over, including coasting to a stop if it loses power or using emergency brakes and/or the engine’s compression if brakes fail. Abundant sensors on the vehicle should be able to diagnose most problems, such as a flat tire. This will enable service personnel to know what equipment or parts to bring, and whether a tow truck is needed.
Also required is the ability to communicate with law enforcement and mechanics arriving on the scene when there is no driver. We expect some standard protocol for these situations to emerge. Perhaps a policeman who finds a truck on the side of the road with no driver in sight knows to look for a specific symbol at a specific spot on the side of the cab to determine if is an AV truck or just abandoned. Perhaps the officer scans a QR code on the truck that connects them to the network control center personnel managing that vehicle to discuss status and what is being done to fix or tow the vehicle. These are just some ideas of what is possible—the actual solutions and standards that emerge will likely be different.
Recent Developments Auger the Coming of Hub-to-hub Driverless Trucking
The prediction that hub-to-hub AV trucking is coming sooner rather than later was bolstered by three recently announced partnerships. Of those three, perhaps most interesting is the Embark/US Express partnership. Embark has developed a hub-to-hub model. They refer to their hubs as ‘transfer points.’ In 2019 they opened transfer points in L.A. and Phoenix and have “conducted hundreds of hauls through these sites, refining required transfer point features and developing process flows for important onsite activities.”
Developing an Operational Playbook
According to the Embark/US Express announcement, “U.S. Xpress’ nationwide terminal network presents an opportunity for a carrier to leverage its existing real estate footprint to support efficient autonomous trucking operations. Through this partnership, Embark and U.S. Xpress will identify priority terminals based on traffic patterns, customer needs, and technical requirements. The companies will start with two terminals in Sunbelt states, creating a clear path to opening a high-volume lane for autonomous hauling.
The two companies plan to co-develop an onsite operations playbook that captures standard processes for when autonomous trucks enter U.S. Xpress properties. Expected solutions will include gate access, onsite vehicle movement, trailer swap procedures, inspections, data and power management, and more. By developing these procedures now, U.S. Xpress will be able to quickly integrate Embark-powered autonomous trucks as a complement to professional truck drivers within its operations.”
The fact that these companies are developing a standardized operational playbook bodes well for scaling up this hub-to-hub approach to many locations. We expect that the next few years will be intense learning, continuing to work out the kinks in how trailers get dropped off and swapped, how drivers get assigned and find their inbound loads, refueling procedures, standardized processes for dealing with vehicle breakdowns, and more. The announcement states U.S. Xpress’ intent to quickly role out this capability, though no exact timeframe was given.
How Soon and How Quickly Will Autonomous Trucking Ramp Up?
Predicting the pace of adoption for autonomous vehicles has turned out to be a perilous endeavor. While many pundits were burned by being far too optimistic a few years ago, it is possible the pendulum has swung the other way to excessively conservative estimates today. We do believe that the challenges for off-highway, public road autonomous vehicles are daunting and will take many years, possibly even decades, to overcome. However, autonomous trucks are already in widespread use in mines around the world (especially in Australia) and starting to be used in ports.
Interstate autonomous trucking is being piloted in many places, especially the Southwest in the U.S. We expect within the next 3-5 years that the volume of goods being moved by autonomous trucks (without a safety driver on board) for the long-haul portion of the trip will increase dramatically from the nascent level it is at today. Much of that volume will initially be in the areas with good weather (no snow/ice, less rain, good visibility), but will migrate to the whole country as sensor technology and all-weather driving models improve. Given the acute, chronic driver shortage, hub-to-hub trucking may prove to be the ‘killer app’ propelling the wider expansion of autonomous truck adoption.