Autonomous trucks: Making the leap to Level 4
By Andre Kranke
I 7 minute read
24/03/2025
Scientists and engineers around the world are working on the next major milestone in the long history of transportation: autonomous driving on public roads.
Since the invention of the wheel some 5,000 years ago, transport vehicles have had a coachman at the reins or, for the last 100 years or so of motorized vehicles, a driver at the wheel. In the future, however, “virtual drivers” will join their ranks.
In recent years, researchers have used artificial intelligence to train specially developed algorithms that are now able to lead a car safely through traffic on public roads. The algorithms are fed with a wide variety of sensor data. For this purpose, modern standard vehicles are equipped with additional sensors such as lidar, radar, cameras, and microphones, all of which provide comprehensive information on traffic events. Based on this data and the behavior patterns learned while navigating traffic, the virtual driver makes its decisions for maneuvering the vehicle on the road autonomously—i.e., without further human support. The best AI algorithms have now achieved a level of safety that’s statistically higher than that of the average comparison group of humans.
The best AI algorithms have now achieved a level of safety that’s statistically higher than that of the average comparison group of humans.
Different levels
Building on these results, a key step in autonomous driving was put into practice in the US last year: as of July 2024, anyone in San Francisco who wants to can hail a self-driving cab. This means that, for the first time in the history of transportation, freely bookable transport from A to B without a human driver in a large public space is now a reality.
To date, only assistance systems that allow Level 3 autonomous driving have been used in production vehicles. In Level 3, the driver can take their hands off the steering wheel for a short time to do other things. However, the driver must be able to retake control of the vehicle at any time.
The division into different levels from 0 to 5 helps to indicate a vehicle’s degree of automation. At Level 0, the human driver drives without any assistance, while Levels 1 and 2 rely on assistance systems and partial automation, such as adaptive cruise control with lane departure warning. Level 5 is fully autonomous, with no human behind the wheel.
Taxicabs as pioneers
Robot cabs from Google subsidiary Waymo in San Francisco have now reached Level 4 of autonomous driving. In a defined area, these vehicles can operate entirely without a human driver. If a situation arises in which the virtual driver is at a loss, it drives to a safe place to park, and a human teleoperator provides the virtual pilot with remote support.
Following a lengthy test phase, Waymo has been operating around 250 cabs as a freely bookable service throughout the city of San Francisco since last summer. Passengers simply hail a cab on the app, get in once it arrives, and then let the all-electric Jaguar’s virtual driver transport them comfortably and safely to their destination. The service is currently being gradually expanded to other US cities. With this live commercial operation, Waymo has proven that autonomous vehicles with virtual drivers will be a reality of transportation in the future.
As part of the “Future Lab” series, results from the Corporate Research & Development department are presented, which were developed in collaboration with specialist departments and branches as well as the DACHSER Enterprise Lab at the Fraunhofer IML and other research and technology partners.
Autonomous truck in development
Start-ups such as Kodiak and Daimler subsidiary Torc Robotics are also working to develop virtual truck drivers in the US. In the next two years, semitrailer trucks will be operating fully autonomously and without safety drivers on a few selected routes on Texas highways. The focus is on hub-to-hub transports between two logistics centers situated close to a highway. The providers are convinced that they will have solved the technological challenges for true Level 4 driving with trucks by then.
Is that realistic? In general, yes, for individual highway transports under defined conditions; this is something that Dachser Corporate Research & Development confirmed during visits to the US sites. However, virtual drivers can’t yet handle certain extreme situations—such as difficult weather conditions with, say, heavy snowfall—well enough to go into actual service.
Major investments
Yet the biggest obstacle to rapid implementation of autonomous driving is the huge initial investment required for the time-consuming training of artificial intelligence. Even for individual highway routes, a great many training kilometers have to be completed, so that establishing a larger number of practical, virtual Level 4 truck drivers for a wide variety of US highways will probably take a whole decade. The attractiveness for investing in this technology and thus the speed of further development in US road freight transport depends primarily on two factors: the worsening driver shortage and rising wage costs, which are already relatively high.
In addition, development in the US could slow down if autonomous vehicles are involved in accidents, which can’t ever be fully avoided, leading to political and legal reactions at the federal or state level.
Global research
Autonomous cars, buses, and trucks are also being developed and tested in Asia, particularly in China. The leading developers here appear to be companies such as Baidu, BYD, and Pony.ai, although Mercedes-Benz reports it is also testing Level 4 driving for passenger cars in Beijing. However, there’s a lack of reliable information on the activities of the various providers in Asia, including details of the results achieved and the safety standards attained.
Europe is currently home only to isolated activities related to autonomous driving. Providers are still a long way from a major deployment of self-driving cars or trucks at Level 4. In particular, there’s a lack of investors for the costly training of AI models. Sooner or later, however, the worsening driver shortage will make it necessary to use this technology in Europe as well, in order to guarantee the efficiency of logistics for the continent as a business location in the future. The question remains as to whether Europe will then have to fall back on solutions from the US or Asia, or whether European companies will play a role in supplying self-driving trucks.
Yet the biggest obstacle to rapid implementation of autonomous driving is the huge initial investment required for the time-consuming training of artificial intelligence.
Use in the logistics industry
Despite all the enthusiasm for future technologies, we must also note that virtual drivers are not a complete replacement for their human colleagues. AI models won’t be able to take over the demanding work of a truck driver in Europe for another decade at the earliest, and then only on selected long-distance hub-to-hub transports. But on many long- and short-distance transports, the diverse tasks of a driver exceed the capabilities of the AI pilot. In addition to controlling the vehicle in complex traffic situations, these tasks include activities such as securing the load, unloading and delivering the goods and, last but not least, personal contact with the sender and recipient of the transported goods. This can’t be done by an AI algorithm in the foreseeable future, so people will continue to play a decisive role in carrying out logistics work. However, they will be supplemented by virtual colleagues who’ll help mitigate the drastic consequences of the demographically induced shortage of skilled workers and drivers.
