Autonomous bullies, ethics and trust

By John Bendel, contributing editor-at-large

By John Bendel, contributing editor-at-large

Just as some kids at school like to bully new students, aggressive human drivers might bully new autonomous vehicles as they begin to show up on the road. The bullies will cut off computer-driven vehicles in traffic and easily nose them out in tight merges. And they’ll laugh as they do it.

Outta my way, wimp

According to the Financial Times of London, that finding emerged from a big study conducted this year in Europe by Goodyear, the tire people, and the London School of Economics. More than 12,000 people were polled on their attitudes toward sharing the road with autonomous vehicles.

OK, the aggressive drivers laughing part wasn’t in the study, but that probably will happen.

The thing is, autonomous vehicles are programmed to err on the side of caution, something aggressive human drivers can use to their advantage. Cut in front of an autonomous car and it will back off. Ignore an alternate merge inching into a single lane, and an autonomous vehicle will not challenge you. There will be no super-tight squeeze play – just a polite "Go ahead, sir. Sorry if I inconvenienced you."

In one focus group, a UK driver who identified himself as an aggressive driver said, "I’ll be overtaking all the time because they’ll be sticking to the rules."

We may not have to wait for autonomous vehicles. There are plenty of semi-autonomous vehicles already out there. They may not drive themselves, but they brake automatically in certain situations. In trucks, there are options from suppliers like Bendix and Meritor Wabco. Similar systems are optional in many high-end cars; they’re now standard on Volvo XC90 SUVs.

Which brings me to a recent ride on a busy Philadelphia street with my son, Tom, behind the wheel. With a lane change looming, he looked over his shoulder and said, "It’s a Volvo; I can probably dive in."

No, he didn’t. Instead, Tom wondered out loud when the radar detection companies will offer a collision-avoidance detector to help in just such situations. "They’ll let you know if the guy next to you has his collision avoidance turned on," he said. "You’ll have a leg up in traffic."

Meanwhile, autonomous vehicle engineers are probably considering the bullying right now. In the comments section of the Financial Times article on the bullies, one contributor suggested drivers of autonomous cars will eventually be able to select from a number of driving-style settings, from "extra cautious" to "aggressive."

Who will an autonomous car choose to kill?

Some very serious people are actually considering that problem. Say your autonomous truck can’t stop in time to avoid a collision and has to decide on an evasive maneuver. Does your truck run over and kill people in a crosswalk or does it crash into concrete and kill you?

The truck in question may not have to be totally autonomous. In July, truck component manufacturers ZF and Wabco demonstrated Evasive Maneuver Assist, or EMA, a system that will know there is insufficient time to avoid a collision by braking alone. Where necessary, EMA will automatically steer a truck away from danger.

You’re probably asking, how can a truck with adaptive cruise and automatic emergency braking find itself going too fast to stop in the first place?

According to a Wabco press release, "critical driving situations may include traffic jams that appear suddenly after curves or when driving on wet, low-friction road surfaces" – in other words, snow and ice.

Wabco and ZF said EMA will steer into an open lane or onto the shoulder, whichever is the safer choice. They did not discuss situations in which there are no safe choices. Maybe EMA can’t tell a Girl Scout troop from a flock of Canada geese, but someday, somewhere it will have to choose between a bad outcome and a worse outcome.

Future robotic systems will very likely know the difference between Girl Scouts and geese. Then how will it decide who it’s going to save and who it’s going to sacrifice? How do we build profound moral choices into self-driving vehicles?

The problem is actually being discussed at the Massachusetts Institute of Technology, among other places.

Platooning will start small

Forget those long trains of trucks we’ve seen in demonstrations.

Well, don’t actually forget them. They may happen one day. But, well before that, we’ll see small truck platoons, most likely the very smallest platoon possible – just two trucks. And we’ll see those paired trucks not very long from now.

That’s the gist of a report from the North American Council for Freight Efficiency. According to the report, two-truck platoons will be relatively easy to organize and will return something like a 4 percent fuel reduction.

"However, real-world factors such as congestion, terrain, weather and road construction will reduce these savings, so fleets will have to estimate this reduction depending upon the routes on which they plan to operate," the report said.

Distance between the platooned trucks will also be a factor. The report says 40 to 50 feet would probably be best, but that may or may not be possible depending on various state laws. Those laws are a challenge and so is public reaction to two-truck platoons when they finally hit the road.

Many of the technical requirements for platooning, particularly two-truck platooning, are already available. They include collision avoidance systems, adaptive cruise control, and mpg optimization systems.

The freight council says other necessary technologies will be on the market within five years. They include the means to exchange operations data between two platooned trucks and software to optimize speed and distance between them for the best fuel savings.

According to the report, "Early platooning adopters will be large, dedicated fleets with numerous trucks operating on a given stretch of highway."

Going beyond two-truck platoons and platooning trucks from different carriers will have to wait a while, according to the freight council. "Platooning operations in North America will be limited to intra-fleet activity until the industry has a better feel for how platooning works in the real world and concerns regarding data transmission between vehicles have been alleviated."

In other words, carriers will have to trust each other.

LL