Long-Haul Trucking With Fewer Drivers

The trucking industry is betting heavily on increasing levels of autonomy and electrification to reduce the cost of moving goods and to overcome persistent problems.

The economics of autonomous driving are compelling, not least of which is an almost perpetual shortage of qualified drivers. But there also are a number of technical hurdles to making this work. On top of the challenges facing the passenger car OEMs, there are issues about real-time adjustments for load shifting and weather, as well as how to ramp up the infrastructure needed to quickly charge large numbers of batteries.

Robert Day, director of automotive partnerships for Arm’s Automotive & IoT line of business, said both the United States and the United Kingdom are short 100,000 qualified truck drivers to transport goods today. That is helping to fuel the narrative around autonomous trucking, whereby trucks will be capable of operating in autonomous mode for freeway driving. So while most experts believe that will take time before Level 4/5 is mainstream in cars, there is growing support that “supervised autonomy” will happen more quickly for the trucking industry.

“In all things related to autonomous driving, the devil is in the details,” said Tom Wong, director of marketing for design IP at Cadence. “What is Level 5 exactly? In laymen’s terms it means given a starting location, a final location, and the vehicle will navigate on its own without human intervention, and will get you safely to the destination. In the ultimate case, there may not even be a steering wheel. Will we get there anytime soon? Today, we are at Level 2 for passenger vehicles. There are two examples of Level 3 that have been recognized by local transport authorities so far, one in Germany (Audi) and one in Japan (Honda). Most automakers have shied away from over-promising Level 5 ability, and instead are focusing their attention on getting to Level 3 and Level 4 sooner rather than later. This is why there is so much chatter on the need for radar versus lidar, in addition to cameras (vision), and the role of sensor fusion.”

But autonomous trucks don’t have to go everywhere that passenger cars can. They can be confined to limited access highways, for example, delivering goods to warehouses/depots alongside those roads.

“We’ve already seen great strides in the heavy trucking industry, because they do have a critical driver from a financial impact perspective, so there’s much more incentive for them to be ahead of the pack,” said Christopher Clark, solutions architect for automotive software and security at Synopsys. “The work that’s been done with semi trucking teaming, as well as in the agricultural industry, has made it clear this is definitely a path they’re going to continue, and they will probably lead as far as regular passenger vehicle capabilities.”

But for the largest automakers, the technology advantage won’t be as evident as some might expect. Auto giants such as Volkswagen, for example, are developing technology in both trucking and passenger cars, and leveraging that technology in both worlds. “In the Volkswagen Group, we have a company called Traton, which is the commercial vehicles division that does heavy trucking,” said Matthew Renna, vice president of e-mobility and innovation at Volkswagen of America. “Having the scale of Volkswagen in batteries and in semiconductors is definitely part of the equation for solving issues in heavy trucking. Obviously there are a lot of different use cases with how the cells are constructed and how the electronics are arranged, but you can certainly see where the technology can cross over and build that scale effect within the Volkswagen Group.”

Competition heats up
Alongside of that, competition is growing in some non-traditional spaces. Clark said a number of startups are working on operational models for after-market or secondary service offerings for heavy trucking.

Consider TuSimple, for example, which is operating commercially in Arizona, New Mexico and Texas with expansion planned. Through its autonomous freight network (AFN), the company is providing capacity as a service to both shippers and carriers. Shippers contract capacity on a per-mile delivery rate using the company’s fleet of autonomous trucks, but carriers also can purchase autonomous trucks and operate them on the AFN, and use TuSimple’s autonomous capacity when needed. Fleet owners also can own and operate their own fleet of autonomous trucks that operate on the AFN.

Fig. 1: The future of trucking? Source: TuSimple

“This is pretty typical of the industry,” Clark said. “Freightliner and other companies provide the heavy truck, and the startup tends to provide the base platform. But it is clear, especially with entrants like Nikola and others, that this is definitely on the way both for EVs and autonomous driving. When we look at adoption from charging stations for typical EVs today for in-home capability, the idea of battery swapping for individual vehicles didn’t really catch on. It may as things progress, as discussed in other countries. But for heavy trucking, you don’t really have that option, so you have to have the heavy trucking industry infrastructure in place. You get the byproduct of the consumer vehicles, so it definitely makes sense — especially when you look at fuel costs and the overall efficiency of those vehicles. They’re looking for half-mile performance increases because that is serious dollars.”

That infrastructure is a critical hurdle at the moment, though. “With trucks, you need the next generation of supercharging technology,” said Sam Geha, CEO of Infineon Technologies’ Memory Solutions group. “By next year, you’re going to start seeing dual charge capabilities. But there will still be infrastructure challenges to solve, and those will take time.”

Geha said there are different motivations behind the electrification and ultimately full automation of trucking. A key one is the impact of diesel engines on global warming. The electrification of vehicles opens the door to more electronics within a truck, and more autonomous features. It also is expected to be a boon for chipmakers, which will be developing the components used in both the electrification of vehicles as well as the ability to supply rapid charging to those vehicles.

But in the case of trucking, there’s also a strong business case to be made for electrification and increasing amounts of autonomy. And too few drivers, with regulations about how long they can spend behind the wheel without a break, reduces the speed at which goods can move and affects the economics of moving them.

“Driver fatigue and lack of sleep for long-haul drivers can cause accidents,” Wong said. “Trucking and logistics/transportation also have very clearly defined routes, which makes it much easier to get ready for Level 5/supervised autonomy. There is also a need to reduce transportation costs because stopping overnight at a truck stop costs money. Autonomy also improves fuel economy and can reduce maintenance costs. This is very unique to large semi-trailers because they are so heavy. In addition, total cost of ownership works in favor of fleets. There are a number of possible solutions. ADAS is very useful for trucking. Driver assist on highways is much simpler to implement — think depot to depot. V2V enables convoying. In addition, the higher cost of lidar is not an impediment to adoption. Big semi-trucks are quite expensive, and incremental costs in using lidar make Level 3 and Level 4 feasible. Further, the cost of sensors is of less concern than in passenger vehicles.”

This calls for a high-speed use case operational design domain (ODD), taken from the urban robotaxi designs. It is primarily forward-looking, with long-distance cameras, lidar, and radar sensors used to detect objects in front of the vehicle.

“This ODD also could include vehicle-to-vehicle communications to enable truck platooning, and due to the long distance nature of the transport, the ODD needs to take into account different and changing environmental conditions (sun, rain, snow), which are often not seen in the geo-fenced robotaxi ODD,” said Arm’s Day. “This means there is a need for many different sensor modalities (lidar, radar, vision) with a heterogeneous compute architecture that will help fuse the data, analyze the environment, and make the appropriate decisions and maneuvers.”

Given the differences in ODDs between trucks and passenger vehicles, does that mean each should have its own vehicle architecture?

“If OEMs take a half step towards next generation automotive platform design, they are significantly different,” said David Fritz, senior director for autonomous and ADAS at Siemens Digital Industries Software. “If they take the big step, where everything is automotive Ethernet — Ethernet connects directly to the actuators, there are central compute capabilities, you put as much near the sensors as you possibly can — then there really isn’t any significant difference other than how algorithms are trained. For example, it takes a lot longer for a big semi truck to accelerate and decelerate than a car, so you have to consider that when you’re putting the vehicle through the test scenarios and verifying that decisions are made properly just because of the differences in mass.”

Another big difference is the number of passengers in a truck versus a car. “In trucking, you’re really concerned about how heavy the load is, or whether the truck is loaded or unloaded,” said Fritz. “That changes dynamically how decisions are made, how much time there is to brake and steer, or whether you should steer around a tractor or just wait behind it. Those are big decisions. But the electronics infrastructure, if they go all the way to next generation, really doesn’t differ that much.”

Moshe Shlisel, CEO of GuardKnox, observed that one European OEM has disconnected its autonomous truck architecture from passenger cars. “This means that they’re not going to be followers of what’s going on in passenger cars and adopt their architecture. They’re going to build their own architecture for trucks’ needs, and that’s very reassuring because if you compare the complexity of passenger cars, sometimes you can find 10 to 15 different internal networks, with everything that comes with them. Right now we have an ECU soup, but in trucks it’s much more mild. There are up to 50 ECUs, and fewer networks. When you go to a different architecture, to address what’s really needed in trucking, we’ll see more and more the zonal architecture approach, and more adoption of high performance computing – also referred to as automotive servers – which are embedded systems based on high performance computing.

GuardKnox is engaged with an OEM and Tier One to build one of these high-performance computers, and looking to a full zonal architecture. “By doing that, you adopt the power of controlling the system remotely,” said Shlisel. “And because you can change it as you grow, you can change the software. If you made some mistakes, you can monitor that on a regular basis, because it is connected. Then, if you implement the cyber security layer from the get go, then you have a remotely controlled and remotely monitored system that can be updated or altered as you need.”

Opportunities abound by leveraging advanced technologies
With the automotive/trucking industry undergoing the biggest transformation since the Model-T, OEMs, tiered players, and startups are looking to the most advanced technologies today. That includes AI, electrification, and cloud computing, as well as other capabilities not seen before in this space.

“This is an interesting industry for third-party providers in the heavy trucking arena,” Synopsys’ Clark noted. “We’re used to seeing other people, not necessarily the manufacturers, providing telemetry information, maintenance information, and many other capabilities, on top of supporting whatever freight they’re carrying. This opens up interesting opportunities. If I’m a short-haul or a long-haul trucker, and this company has brought this interesting battery capability that’s part of the trailer and it just plugs into my truck, now I have extended 300 to 400 miles. That’s really compelling. There are a lot of open innovation areas in the heavy trucking industry for existing and new organizations to participate in.”

Interestingly, when Intel-owned Mobileye started out with its technology, the first market was aftermarket selling to trucks. “The idea was that — and it was true — if you purchase this Mobileye system and install that in your fleet of trucks, your insurance would go down,” said Kurt Shuler, vice president of marketing at Arteris IP. “That was the value proposition, and it retrofitted these capabilities onto these trucks. There’s a lot of liability with these trucks when there’s an accident. That’s why companies like FedEx Ground and Amazon, instead of doing like UPS and saying, ‘These trucks are part of our fleet. We own them,’ they’re independent contractors. And so if your Amazon delivery or your FedEx Ground delivery crashes at FedEx, it’s the responsibility of the contractor because of the liability around that. There are incentives in the economics for these kinds of things to add the capabilities.”

Autonomous trucks can have a big impact on cost competitiveness, too. “In long-haul trucking, because the rules are such that there are limits to how many hours the truck driver can drive even, even if a truck is not 100% autonomous, if they can extend that time to make a load cheaper or longer, that’s a benefit,” said Shuler. “Anytime a truck is sitting still, it’s not making money. If they could make better use of the existing infrastructure, that’s an easy thing to decide. If you’ve got an expensive capital asset and you can get 2X the work out of it, then going across the whole country might be more feasible. If you’re driving across Kansas or Nebraska, it’s fine, but once you get into St. Louis, maybe the system can’t handle it, so the driver has to grab the wheel. The technology makes it very attractive to do that, and it ends up being a testing ground for these things once you get to the full autonomous drive.”

Further, an important target for EV designs is to help improve battery life and power management. Embedded FPGA technology can help with this.

“By integrating high performance embedded FPGA technology into an ASIC, developers can integrate more functionality, increase performance, increase flexibility and also save power. Certain architecture can be used to reduce soft-error rates through using triple redundancy and using other techniques for high reliability applications,” explained Andy Jaros, vice president of sales at Flex Logix.

Still, it’s difficult to pinpoint how long it’s going to be before fully autonomous trucks are used en masse. The technology may be there, but human beings are still reluctant to embrace it completely.

One tipping point, at least in the trucking industry, may be the truck driver shortage. “Not many people want to do that job,” Shuler said. “It doesn’t pay that much money. There’s always a shortage, which means the fleet companies aren’t paying enough to attract people to do it. Who wants to live on the road in a truck? Who wants to do that for a living? And because there’s always a shortage of people who want to be commercial long-haul truckers, it seems to follow that’s going to be one of the first places where this technology is applied. At first, it seems likely that it’s going to extend the working day of these truckers. It’s going to get good enough to where, in most cases, it can take over if the truck driver falls asleep.”

Another area with a lot of potential is in fleet management, he said. “This would be on a local level where routes are highly optimized by companies like UPS, FedEx, and the like, because you still need to have a human being to drop off the packages or pick them up currently. But could you extend the driver’s day once the technology is good enough? That’ll come after the long-haul trucking. Once the technology is good enough, initially you start extending the driver’s day, and then Rosie the robot can deliver the package.”

Fritz expects most of the benefits that will be realized for the next decade of full autonomous will be tempered by the fact that there still needs to be someone in the cab, so the real value comes from not having to stop for breaks, people not falling asleep at the wheel, being more efficient about navigating through high traffic situations. “If you have an autonomous vehicle, even if there’s a driver, you could save 40% of total driving time simply by skipping restroom breaks, and stopping for dinner and all those things. Even though there will be platooning and things like that, I still see some sort of human involvement. So it might not be a total redesign of the cab. There still needs to be the ability to kind of oversee and override, just because there’s so much more riding on it. In a passenger vehicle you may have passengers in the vehicle, but in a truck you may have $200,000 worth of product in the trailer. It’s going to take a bit longer for the trucking industry to just say, ‘Yeah, let that thing fly on its own, and if it crashes into a couple of cars, no big deal. We’ll have legal take care of it.’ I don’t think that’s the direction it’s going to go.”

As trucking moves closer to autonomy, other related issues come more into focus including the ability to sense the load in the trailer shifting, Fritz said. “There are an awful lot of loads in the back of those trucks that can shift even a few inches. So if it’s not a balanced load, then somehow the system has to recognize that and adjust accordingly.

Another dimension related to autonomy and the load shifting are environmental conditions.

“If the road is wet, what is your braking distance? Then, other dimension on top of that is, once I hit the brake, if I go sideways because the road is wet, the load is going to shift. How does that change the decision I should have made before that happened? Also, what happens when the load weight changes? How does the system make decisions dynamically? The truck might have five stops from point A to point B. Each time something is pulled, and/or something is put in, the weight is different, the distribution is different, so things have to be recalculated on the fly. That’s extra complexity that passenger cars don’t have to deal with,” he explained.

This points to additional electronics inside the trailer, including sensors and gyros, to provide data about the load weight and shift.

Yet another aspect of this is the concept of momentum, and how momentum impacts the vehicle. “If it’s an unbalanced load or a load that shifts, then the momentum vector changes. A lot of this can be sensed. The question is whether you can sense it in enough time to make the proper decision in these corner case scenarios, like, it’s snowing,” Fritz said.

More advanced features including in-cabin vision solutions such as distracted driver monitoring, human presence detection, and driver identification must be addressed going forward. Here, reconfigurable computing approach are being considered given that a single chip can be used for several applications without incurring multiple high ASIC chip development costs. 

“A single reconfigurable device can offer high performance at low power for more than one use case while also offering the ability to be reprogrammed for different algorithms remotely via over the air updates, thus supporting both bug fixes and feature enhancements over time. This addresses the needs of EV OEM manufacturers not only save on power consumption but also reduce the need to source additional parts and reduce the size of their modules which can be critical factors for product success,” Dana McCarty, vice president of sales and marketing at Flex Logix said.

Conclusion
When it comes to the whole spectrum of evolution towards full autonomy between trucks and passenger vehicles, Fritz noted there’s also everything from tractors and combines and other agricultural equipment, which are quickly becoming 100% autonomous because the risk is relatively low if there’s a fault.

“At the other end of that spectrum, with passenger vehicles and trucks, there’s still a lot of interest and concern. That being the case, the emphasis on safety, security, and reliability for trucking, as well as passenger vehicles, are pretty much the same. From the insurance company’s perspective, they’re going to have a say in this. You can see how the burden of insurance is going to dictate what platforms are, or are not allowed to happen. This is coalescing in passenger vehicles and the trucking industry together. Yes, there are savings, but you’re not going to be relieved of all of the safety and security and reliability issues that passenger vehicles have. Given that the vast majority of trucking is still at Level 2 and Level 3, whereas passenger cars are starting to hit their stride at Level 4 and Level 5, the timeframes are going to be different for autonomous deployment.”