In October 2018, the US Department of Transportation (DOT) published Preparing for the Future of Transportation: Automated Vehicle 3.0 (AV3.0), building on previous guidance to outline its vision for transport beyond passenger cars. It acknowledges the potential for automation to transform freight transportation, and states its goals moving forward will be to modernise regulations, and encourage a consistent regulatory environment across constituent states. For example, it suggests certain states could modify laws concerning vehicle following distance, as these could hamper platooning applications.
Such commitments to eradicating regulatory barriers are exciting for autonomous truck developers such as TuSimple. Chuck Price, Chief Product Officer at the US and China-based company, says regulatory matters now represent perhaps the biggest obstacle to the self-driving roll-out across the US. “Right now, the regulatory environment across the different states is a patchwork,” he explains. “TuSimple is hoping for a 50-state solution, and the launch of the AV3.0 initiative by the DOT gives us hope. This, along with other initiatives, will reduce the burden of getting the technology accepted.”
This is not to say that TuSimple has already perfected its offering. The mild tempering of expectations for self-driving that has followed in the wake of the self-driving Uber crash in March 2018 has illustrated that technical challenges remain for developers, such as object detection in all conditions and dealing with extreme but safety-critical corner cases. But whilst ten years ago the thought of self-driving semis would have made many scoff, today the self-driving truck appears tantalisingly close, and the the DOT’s AV3.0 means those with the final say—the government—are fully on board.
“TuSimple has moved beyond the research phase,” says Price. “This is no longer rocket science, but a matter of engineering.” The figures appear to back this up. TuSimple has recently announced unicorn status, following its Series D funding round. Funds will grow the company’s fleet to 50 trucks by June 2019, taking testing out of Arizona and into Texas. Its strategy is two-fold: the company is running its own fleet services, making up to five deliveries a day for 12 customers. At the same time, it is working with truck-makers and Tier 1s on joint production programmes.
However, exactly how the self-driving future will play out remains to be seen. Different companies are proposing different visions, and cultivating different ideas for applications. For TuSimple, little short of full autonomy will suffice.
My way or the highway
Ike Robotics is one of the latest arrivals to the self-driving truck market. Its proposition is to ‘de-scope’ the autonomous challenge by strictly limiting self-driving applications that use its technology to the highway. Chief Executive Alden Woodrow, the former lead for Uber’s self-driving truck programme prior to its closure, has previously told Wired magazine that trucks fitted with Ike’s technology “won’t take a single right turn off the highway”. All other driving tasks will be handled by a driver.
It makes sense to some: the highway is an infinitely less complex environment than a built up urban centre, with long straight roads, largely clear markings and virtually no pedestrians nor traffic light infrastructure. What’s more, freight trucks will spend much of their driving lives on the highway. Ike’s model envisions transfer hubs, directly off the highway, much like a service station. Here, humans can take the wheel, or swap out the self-driving cab for their own, and complete the journey.
The problem with the transfer hub model, says Price, is that it would require an infrastructure overhaul. “A transfer hub is basically a giant parking lot right next to the highway,” says Price. “The idea sounds good in theory, but let’s imagine the process being executed for one hundred trucks at a time. Straight away you need 100 acres of land, security needs to be built up, and then there’s the task of managing this process for large numbers of trucks. It starts to look like a port, and we know what port operations look like: they’re nightmares.”
What’s more, anything less than full autonomy, he argues, doesn’t solve one of the main challenges facing the trucking industry today: a chronic driver shortage. Today’s driver population is already ageing, with the average age of US drivers up to around 55 years. Meanwhile, freight transport demand continues to grow dramatically. “You will not find depots on the highway,” says Price. “Sometimes, they’re several miles off. And so a vehicle that cannot operate autonomously off the highway doesn’t solve the driver shortage issue for our customers.
“Fleets need a solution that adds significant value,” he continues, “or there’s no point in adopting it. If instead what it’s doing is adding significant burden without a commensurate dramatic reduction in cost, it’s not worthwhile to them.”
The bar is high
Of course, designing a solution which can navigate off-the-highway is a far more challenging technical feat. Key to enabling depot travel will be additional capabilities, says Price, such as vehicle monitoring and recovery services. Other value-add services, such as autonomous refuelling, will form an important part of the offer.
The other crucial challenge is sensors. TuSimple favours cameras, which Price argues have a longer range than LiDAR: whilst the latter can only achieve effective distance of 100 metres, he says, cameras can provide ten times that at a thousand metres, enabling trucks to reach full speed on the highway. Should a 40-tonne articulated truck encounter a braking event with only a hundred metres’ notice on the highway, it may already be too late. But cameras are often cited as being susceptible to failure, whether through weather or otherwise. Price believes TuSimple has found a workaround: “We have trained our AI system to be able to see through inclement conditions such as rain and dust,” he says. “There is also the secondary radar system. Sensor technology is advancing quite rapidly, and the ability to be able to see in extremely low light conditions is really improving.”
In short, Price is confident that the technology is almost there. Much remains to be done, not least further validation and additional testing miles for the full confidence of regulators. Whilst AV3.0 may demonstrate enthusiasm on their part, it will be a foolish government that rushes the self-driving agenda before it is fully ready, as perhaps demonstrated last March in Arizona.
This article appeared in the Q2 2019 issue of M:bility | Magazine. Follow this link to download the full issue.
