AI cars face tricky culture clash in understanding human behaviour
Image credit: image source
Before human beings can be removed from the driver’s seat, it is necessary to know how people use their cars and, more importantly, how they communicate with others on the road.
For manufacturers of self-driving vehicles, the technology is relatively straightforward – radars, cameras, artificial intelligence (AI), advanced digital mapping and so on. Understanding the ‘unofficial rules of the road’ from one country to the next is the challenge.
Compare the notoriously chaotic and often fatal traffic in Ho Chi Minh City, Vietnam, to the more orderly road users of Munich, Germany – they are vastly different.
It’s said that autonomous vehicles will be everywhere in towns and cities from 2020, significantly reducing accidents by taking people out of the driving equation. Yet how can they be adapted to meet the needs of many countries with widely differing road conditions and cultures?
Bernhard Weidemann, spokesperson for autonomous driving at Mercedes-Benz, says manufacturers will eventually address this with diverse model and package options, but don’t have clear distinctions yet.
First, they are working to overcome problems identified from trials in Western countries, like how can self-driving cars replicate human interactions and be accepted on the roads?
In most countries, drivers and pedestrians use eye contact, hand gestures and other expressions to communicate intentions.
Autonomous vehicles, however, cannot recognise these signs. This became apparent during a long-distance test journey in 2013 by Mercedes-Benz in Germany with its S500 Intelligent Drive technology. As the car approached a pedestrian crossing, it stopped for an elderly lady. The woman waved the vehicle on as she wanted to cross slowly, but the system could not recognise the gesture. The safety driver had to intervene.
For the Mercedes-Benz F015 Luxury in Motion concept model, unveiled at the CES 2016 trade show, designers tried to mitigate this problem by putting blue lights on the exterior to inform other road users that the car is computer-controlled.
“A pedestrian can then pass or turn away from the crossing so the AI can continue – this is crucial to the success of self-driving vehicles, as it will help people understand and accept them,” says Weidemann.
Nissan has hired an anthropologist to help designers with such issues. The company’s IDS Concept vehicle, a vision for 2020, also uses external lights and displays to convey its awareness and intentions to people, flashing messages like ‘After You’ when giving way.
Truly autonomous vehicles will need to interact with human drivers, but again, will they have the capability?
As well as flashing indicators to communicate their intentions, people also use all sorts of physical and personal gestures. If these interactions don’t go smoothly, tensions can escalate.
“People will be unsure how to interact with autonomous vehicles,” says Chris Tennant from the Department of Psychological and Behavioural Science at London School of Economics (LSE).
Tennant and his colleagues conducted a study asking people how they feel about interacting with self-driving motors. It included 48 focus-group participants and an online survey, answered by approximately 12,000 European respondents.
He describes a scenario: an autonomous vehicle needs to overtake a parked delivery truck on a busy two-lane urban road. When there is a gap it must speedily pull out, going slightly into the other lane, and around the obstruction.
“In that situation, some people think the car will get stuck there all day because it will see an obstacle and if there is a lot of traffic, it will stop,” Tennant says. “Some commentators were clear that they would just nip right round and take advantage of the situation.”
This could be a bigger problem in some countries than others. “In Istanbul, if you don’t push into the traffic you will get rammed from behind, because they are already anticipating that is what you will do,” he says. “That doesn’t sound like an environment that autonomous vehicles can function in.”
A possible solution for such scenarios, including roundabouts with no traffic lights, says Weidemann, is additional sensors on the vehicle’s sides. This means that surrounding traffic is detected earlier and the car can act accordingly.
There are lots of issues to consider, such as fallen trees or traffic accidents, which would require breaking conventional road rules to avoid the obstacle.
Nissan says its Seamless Autonomous Mobility technology could use cloud connectivity. This tech would contact a command centre in situations when a vehicle ‘should not attempt to negotiate the problem by itself’.
A human operator would then access the car’s sensors and cameras and use a remote map to establish a safe course, ‘painting’ a virtual lane around the hazard for it to follow.
One benefit of autonomy in vehicles is that it gives passengers the freedom to concentrate on other matters instead of the road.
“Giving business people more flexible time for work will be very important for progress and social acceptance of self-driving cars over the next five years,” says Richard Viereckl, partner with PwC Strategy& Germany.
The Mercedes-Benz F015 Luxury in Motion can transform into a mobile office by rotating the front seats to face other occupants, who can then use the interactive displays to view information together.
Yet Mercedes-Benz design chief Gorden Wagener believes people will always want the option of driving themselves, and most design concepts still have a steering wheel.
The VW I-D Buzz is a futuristic version of the classic VW Camper. Its steering wheel can be gently pushed and will disappear into the dashboard. When the autonomous mode is on, the driver can swivel their seat around to socialise.
In theory, users will relax and enjoy the ride, but this is only likely to happen when people begin to trust the technology.
When participants of a Goodyear survey were asked what they would do with this free time, 82 per cent said they would spend it watching the road, even though it’s unnecessary.
Perhaps this lack of trust is why manufacturers see the need to incorporate Siri-like communication capabilities.
The Nissan IDS Concept AI communicates like a reassuring ‘attentive partner’ according to the company, updating the user on traffic conditions, their schedule and personal interests.
Toyota’s Concept-i, revealed at this year’s CES, imagines driving in 2030. Its key feature is an on-board ‘Super-Siri’ named ‘Yui’ represented by a flashing blue circle that can be used to operate every function of the car simply by asking. Yui learns about its user, assesses mood and anticipates their needs.
Autonomy is already installed incrementally into cars. All Tesla vehicles are operating with self-driving capabilities as of October 2016, and most high-end automobiles have similar systems.
In Japan, Nissan’s Serena model has ProPILOT software for autonomous control on the highway in single-lane traffic. Multiple-lane driving is coming in 2018.
Weidemann says Mercedes-Benz foresees this technology being adopted in two ways: motors could use deep learning to improve their assistance systems’ ability to handle different situations until the user can relinquish control, or autonomy happens immediately.
This would only happen in designated areas at first, such as carparks and selected areas of the city, then expands.
This all depends on what market the vehicle is sold in. Nothing is certain, but industry experts anticipate machine-driven motors in the US first for business people, then Japan, modern Middle Eastern cities with lots of financial capital, as well as designated spaces in Europe.
For each country, relevant legislation will have to be created, a government licence granted and the car’s human driver interface – its brain – tailored to the market. Above all, they must have reached social acceptance.
Experts agree that first-case use of autonomous systems will almost certainly be in taxis or long-haul logistics. These applications are thought to be the most commercially viable, cost-effective and easily managed.
Yet it’s unknown how this might be undertaken and managed in different places.
“Right now, we have different traffic systems and infrastructure. In some places, such as India and Vietnam, there are no road rules,” says Viereckl.
He imagines future autonomous vehicles across the world will have their own lanes and follow their own traffic guidance, but he can’t be sure. “The critical question now is: will infrastructure change in each country, or will we need to variant the cars completely,” he adds.
Most likely, the answer lies somewhere in between.