Yellow rubber ducks could be one way of avoiding deadly accidents with self-driving cars, such as the one in which a
Tesla Model S was involved in Florida earlier this year. Commonly found in bathtubs, the little toys also populate the streets of Duckietown, an advanced autonomy class taught at MIT.
Built and programmed by students, the 50 robot-controlled ducks have to recognize traffic lights, road signs and other obstacles that challenge auto engineers and drivers in the real world. Each robot taxi is equipped with a single camera and navigates the model roads without using pre-programmed maps. In a playful yet clever way, the ducks are there to test how self-driving cars might interact on public roads in the future.
“One of the major challenges for autonomous driving research is the safety and logistics associated with this work,” says
Liam Paull, who leads the project. “If we can develop a smaller, safer, cheaper platform where the same algorithms work as for real cars, we can carry out research more efficiently.”
In the meantime,
Google is running a self-driving car project with a testing program in four US cities using cars that have autonomously driven over 1.5 million miles.
Three types of sensors enable the prototype to “see” its surroundings, even at night, to predict the behavior of what it sees and to make decisions about how to react by comparing what’s happening in real time with its models. Cars are taught to detect unfamiliar objects or other road users, and are given practice on how to respond.
LET ME GOOGLE THAT CAR
The challenge is to translate the complex technology behind autonomous cars into a product that is accessible to user groups such as the elderly, says Timo Möller, automotive consultant at McKinsey, Germany. “It will have to be a case of a senior citizen giving the car simple instructions, such as ‘take me to the doctor’s.’ The car might then ask if the driver wants the fast or the slow route. It should be as simple as getting onto a bus.”
Self-driving robo taxis might be an attractive form of transport for the elderly. These fully autonomous vehicles, currently under development for commuting short distances, could be ordered for that weekly trip to the doctor’s surgery, for example.
Overall, autonomous vehicles could offer individual mobility for the elderly despite physical, health and other handicaps that may come with old age, says Sven Beiker, of McKinsey, a former executive director of
CARS, the Center for Automotive Research at Stanford University. “Mobility is a human need, on a par with a desire to communicate. Being mobile is key to having an active lifestyle and to leading the life you want: if a car takes over the driving, then it doesn’t matter if you are able to drive or not.”
Being mobile is key to having an active lifestyle Sven Beiker
More than 90% of road accidents include human error as a contributing factor but introducing self-driving cars will not necessarily prevent all those accidents, says Beiker. “There is this huge expectation that once we have this technology, we have a high chance of getting rid of a large share of accidents based on human error. Yet how good we can get overall is still not decided.”
One problem is that computers can’t think outside the box and are ill prepared for non-standard traffic situations. A human is experience and intuition based, cutting corners here and there, explains Beiker. A computer doesn’t know how to react to a situation that hasn’t been programmed.
Google experienced its first collision in February, when one of its self-driving Lexus SUVs drove into the side of a bus. The question of safety remains – until it might be answered by a duck.