Sensing Changes in Autonomous Trucks

October 1, 2018

Terry Costlow

Driver shortages and rising demand are fueling interest in technologies that let commercial trucks navigate with minimal to no human input. Many of the sensing systems being designed into trucks are borrowed from passenger cars, but commercial vehicles have many different requirements.

Radar, cameras, lidar and control technologies for automated vehicles are all advancing quickly, giving commercial vehicle engineers a number of tools to work with. Highway speeds are largely universal, but many factors change when these electronic systems migrate from cars to trucks. The systems and their software will have to be honed to operate with vastly different parameters.

“The sheer size and weight differences between a passenger car and a commercial vehicle at 80,000 pounds mean the actuation systems for acceleration, braking and steering are of a different magnitude as well,” said Derek Rotz, Director of Advanced Engineering for Daimler Trucks North America. “Stopping distances vary from cars to trucks, so the need for earlier warning of objects becomes more critical. Trucks operate under a wide array of conditions, from fully loaded to empty. These differences cause vehicle dynamics to vary, which must also be considered.”

The vagaries of commercial environments also pose major challenges for design teams. Over their operating lifetimes, tractors will haul loads of widely divergent weights and sizes, factors which impact braking and steering.

Additionally, most autonomous programs call for 360-degree views around the vehicle, so sensors will have to be mounted on trailers. The electronic links between trucks and cargo carriers need to facilitate frequent configuration changes. Connections and communications between tractor and trailer will work best if they’re simple and universal.

“Tractor-trailer combinations are not a given during equipment lifetimes — there will be a lot of swaps,” said Dan Williams, Director, ADAS and Autonomy, at ZF Commercial Vehicle Technologies. “That creates a demand for standardization.”

Several sensors

A range of sensors will be needed to provide the input needed by autonomous control systems. Though the number of radars, cameras and lidar and their locations will vary widely depending on vehicle types and loads, all of them will be needed regardless of how they are configured. The technologies all have different traits, so using a mix provides more certainty about what’s in the vehicle’s purview.

“We feel it is critical that a comprehensive approach is taken as each of these types of sensors play a specific role,” Rotz said. “When deployed together, they can collectively improve overall safety performance.”

Multiple sensors complement each other, providing information that can be combined to improve identification and avoid false positives. Using many sensors also provides redundancy for a range of failures. Radar and lidar can provide results when a camera can’t see through snow, for example, letting the systems understand the surrounding environment.

“Automation requires more and more situational awareness and redundancy,” Williams said. “We need a wider field of view and redundancy, to be robust when one sensor fails. That redundancy includes having different sensor technologies.”

Lidar is the new technology in this mix. Solid-state lidar systems are still in their infancy, with several suppliers vying for what’s expected to be a huge automotive market. These cameras add lasers to work much like radar, adding distance capabilities to identification skills.

“Lidar can provide very accurate measurements, especially in comparison to radar,” said Chris Woodard, Business Development Manager for Autonomous Machines, Danfoss Power Solutions. “Rain, snow and dust can all take a toll on lidar, where it doesn’t bother radar at all. From a safety standpoint, lidar is a great solution and complementary sensor to other available technologies.”

Off-the-vehicle input

Sensors are typically mounted on the vehicle, but in some commercial environments it may be more beneficial to install sensors in the infrastructure. In warehouse docks or a mining site, fixed infrastructure sensors can determine where vehicles are. Locational data can be transmitted to the vehicle, where it can be treated like any other sensor input. That can keep costs down.

“Automated behavior in freight yards, which extends to outside operations like pit to port, has very different concerns because operating environments are more controlled, vehicles are not exposed to the general public,” Williams said. “One great advantage is that you can put instrumentation in the freight yard so the back of every trailer doesn’t have to be covered with sensors that cost more than the trailer and its freight.”

Remote controls are another way to gain the benefits of autonomy without making the investment needed for fully autonomous vehicles. In environments like mining and agriculture, a single remote operator can monitor a few semi-autonomous vehicles, taking control mainly when systems aren’t able to make a decision based on the sensor input. This technology may help companies take steps towards autonomy while developers are working to bring driverless vehicles to mainstream applications.

“We see remote control systems as an important piece on the path to full autonomy and will play a critical role in developing an overall machine control package,” Woodard said. “In the future, I can see this technology evolving to more ‘tele-operation’ applications, where operators can control machines through a long-range radio link or over an internet connection.”

---