Driverless Trucks Are on Texas Highways — but Not Taking Over

By Olivia Hart

Aurora is running large freight trucks with no driver on I-45 between Dallas and Houston right now — and that flat, straight, carefully chosen corridor is exactly the point. Driverless trucking exists today only on roads stripped of complications, according to Shannon Roberts, an associate professor of industrial engineering at UMass Amherst who studies human factors and transportation safety. Widespread use, she argues, remains far off.

What can automated trucks actually do?

A fully automated truck — one that can drive an interstate in good weather without human intervention — combines AI, machine learning, computer vision, pattern recognition and adaptive decision-making. Torc, one developer, uses AI to merge data from multiple sensors into models that predict truck performance.

The claimed benefits are concrete. Data from modern rigs already equipped with automatic emergency braking and lane-keeping assistance suggests driverless trucks could cut both the number and severity of crashes. Optimizing acceleration and braking, and running trucks in close convoys, could save fuel and cost. Trucks able to run nearly 24 hours a day would need less parking and would idle less at truck stops, cutting emissions.

Automation might even ease the industry's driver shortage — ironically, by making driving less stressful and less physically demanding, which research links to better job satisfaction. Every promised benefit, though, rests on favorable conditions the real road network does not reliably supply.

Why aren't they everywhere yet?

The main obstacle is variability. Snow and rain block the sensors, radars and cameras automated trucks depend on to "see" the roadway — a problem for all automated vehicles, not just trucks. Waymo, the driverless taxi operator, runs its cars only in areas that get no snow.

Construction zones are worse. Orange cones, narrow lanes and missing lane lines defeat the technology, and driverless car companies still try to avoid these areas. One odd research finding: human drivers crash into driverless cars at a higher rate than into cars piloted by people.

The test routes confirm the limits. Aurora picked I-45, a relatively flat, straight highway. Rival Kodiak's robotrucks deliver sand to industrial sites in Texas over roads few people use. Almost no automated trucks are being tested in chaotic urban environments with pedestrians, bicyclists and traffic lights.

The market has already culled the field. In 2023, Embark, TuSimple and Locomotion shut down. Waymo killed its trucking program to focus solely on cars. Other companies narrowed their ambitions to a few dedicated routes between two large cities or manufacturing sites.

Will drivers stay in the cab?

For years to come, yes — in Roberts's view, AI-controlled trucks will still require a driver. The realistic model is a division of labor. A human drives the "first mile" through crowded warehouse grounds and local streets to the highway, then releases the wheel, pedals and brake while automation handles the interstate, then takes back control for the "last mile" to the destination.

Research by Roberts and her colleagues shows human-machine interfaces will be decisive. A screen or head-up display would tell the driver how the automation is performing; verbal messages or a vibrating seat would signal when the system cannot handle a situation.

She flags two biases developers must manage. Automation bias could produce drivers who trust the truck's decision-making so much they defer to it. Algorithmic bias could cause sensors to recognize certain groups of road users more than others — detecting, say, only pedestrians of a certain size or shape.

The long-term vision is a cab with no human at all, partly to lower costs. But Roberts does not see driverless trucks as practical at large scale in the near future; manufacturers, technology developers, regulators and drivers will have to work as a team to capture the benefits. In plain terms: the driver's seat stays occupied.