Teaching Volvos how to change lanes autonomously
A Volvo that can overtake all by itself – how will that work?
“It’s like your sat nav, just more elaborate. In the future you will set your destination and your car will calculate a route – just as it does today – but then it’ll look at the route and decide in advance which lanes to position itself in as it drives along. After that, as it drives along, it will have to decide when exactly to start a lane change – which gap to move into, how much to accelerate or slow down. So there are lots of decisions to take by the algorithms we develop.”
Can the car change its mind if circumstances change?
“Yes. For example, the car might abort a lane change if a car behind suddenly starts accelerating – but only up to a point. Once you’re over the lane marker, you don’t abort the maneuver.”
What if the car gets into an ethical dilemma? There might be an obstacle forcing the car to swerve, but there is oncoming traffic?
“Our approach is to take precautionary measures all the time so we don't end up in situations we can’t handle. Let’s say we pass a car standing on side of road. If a pedestrian were to jump out, we’d have no chance to react. That’s why we take measures in advance. If we can’t see what’s behind the car as we drive past, we slow down enough so that we can react if necessary. We also make sure there is enough of a margin to move sideways if need be.”
Apart from lane changes, which other scenarios are you looking at?
“There’s the decision to take the car to a safe stop. If you ever have a fault the car needs to be able to handle this on its own – we never rely on the driver taking over. So we need to be able to maneuver the car to a safe [stopping] location.”
When will this technology be available to customers?
“The first customers will be able to test e.g. semi autonomous lane changes in the pilot starting next year. The first cars in the pilot, will be highly capable autonomous cars that need to be supervised by the driver. Once on the AD route, the cars will be able to handle most driving scenarios without operation by the driver, but he or she still needs to supervise and is still responsible for the car. Then we will need around two more years to develop the technology further and prove that this system is absolutely safe. We hope to have fully autonomous cars in production around 2020, although there’s lots of uncertainty.”
Apart from giving people back some of their time on commutes and motorways, what are the benefits of this technology?
“Safety. Self-driving cars are much more predictable than humans.”
And what is your greatest challenge?
“The greatest challenge is humans in traffic – how do we tackle the uncertainty of how other road users might behave? We have to make sure that whatever reasonable manoeuvre another car makes, we stay safe – but without being over careful. The other challenge is to prove conclusively that this system is safe.”
How will you ensure that the technology is safe?
The critical part of the validation is to ensure that the vehicle can handle all exceptional situations that may occur on the road. Exceptional situations may occur in terms of traffic scenarios, weather conditions or system faults. To make sure that the system can handle this we will collect large amounts of measurement data from real world driving, but we also will simulate this situations used computer clusters.
In other words, self-driving cars will be much safer than a human driver?
“Yes – much safer.”