What happens when things go wrong on self-driving trucks

Kodiak Robotics has been practicing the maneuver since last year. Last week, the company held a public demonstration, cutting an Ethernet cable that is part of its self-governing system, and showcasing a video of a Class 8 truck that then stretched to its shoulder along Interstate 45 near Dallas.

A company from Mountain View, California, which has built its systems to check for faults 10 times per second, has announced a test first in the industry. Competitors say they conducted similar testing.

Waymo, who has experience in both robotics and trucking applications, says he regularly tests his trucks in so-called minimum risk scenarios on closed tracks, planning to get trucks to shoulder or off the road at close range. ride.

Another competitor, Aurora, expects to showcase this ability on the highway in the third quarter of this year. The company says it is now improving traction on the shoulder in simulations, on test tracks and in limited conditions on public roads.

In Aurora’s first-quarter earnings report this month, CEO and co-founder Chris Urmson stressed the importance of trouble-free operations, saying improving them is a central element of the company’s plans before beginning of commercial service in 2023.

“More often than not, we talk about these sexual problems of machine learning, computer vision, which are obviously important,” Urmson said. “To actually have a commercially viable product, you have to deal with what happens when the product somehow breaks down.”

Although companies are now highlighting fail-safe issues, they have long been considered.

For Aurora, this included working with partner Volvo Autonomous Solutions integrate their self-government systems into the Volvo VNL long-haul truck and develop backup systems for braking, steering and driving.

Similarly, when Burnett launched Kodiak in July 2018, he didn’t want to just prepare for a stand-alone test in which he could count on a human security driver as a backup. He wanted to include backup conditions to begin with.

“Driver, take over” is a pretty logical and sensible way to design the early stages, because you don’t have to worry about handling the system in a degraded way, “he said. “But it’s a short approach. We said that from day one we were going to calculate both the nominal path and the alternate path, and in parallel created the complexity of these systems.”

Kodiak’s design divides its computing mechanism into two different systems: one that handles road situation understanding, decision-making and other primary tasks, and the other that is certified according to the industry’s strictest level of automotive safety integrity (ASIL-D). He can take control when a fault is detected and go to the side of the road. Waymo and Aurora have similar primary and backup computing architectures.

Mapping can be an industry challenge because making accurate high-definition maps is more difficult. This is not a problem for the Kodiak system, which relies less on high-definition maps and more on real-time neighborhood output.

Roads themselves can be accidental situations littered with debris, roadblocks and other vehicles that are either stopped or have uncertain trajectories. However, the shoulder is almost always preferable to stopping in the lane.

“We consider it a last resort,” Burnett said. “It’s usually not considered the safest option in most circumstances. Whenever there is free shoulder and access to it, our truck will try to do it.”

Once parked on the shoulder, there is another problem. Federal Trust Safety Regulations prescribes that drivers of vehicles stopped on the highway or sidewalk must place warning triangles or flares within 10 minutes of stopping.

A Kodiak spokesman said the company is working with regulators and others to find a solution to a problem that remains unclear across the industry. This will need to be addressed before the start of commercial deployment next year.

At the moment, the roadside is considered a milestone.