How to humanize the autonomous car

So far in our series ‘3 Steps to automation’ we’ve looked at how HD maps and Live Roads will be essential to ensuring the feasibility of autonomous cars. However, we believe there is one more very important factor to make the driverless car a mainstream success and that is the need to humanize the autonomous driving experience.

Theoretically, a car could drive itself at a speed that feels too fast and still be perfectly safe – but it may not necessarily make passengers feel relaxed or comfortable. By analyzing long-term location-based driver behavior and understanding how humans drive, it is possible to make the autonomous driving experience more familiar and comfortable.

Forming the third component of the HERE automated driving cloud, Humanized Driving is going one step further and not only allowing your vehicle to mimic human driving style, but your personal style.

What’s a comfortable driving speed?

When it comes to comfort, one size does not fit all. A comfortable ride is a matter of perception that depends on a number of factors.

“These include, but are not limited to, road geometry and width; oncoming traffic; oncoming traffic at night; weather conditions; the condition of the road surface; roadside objects such as buildings or trees obstructing the view through bends; social and cultural driving norms; and personal preferences,” says Ogi Redzic, Senior Vice President, Connected Driving at HERE.

By analyzing probe data from non-automated cars already on the road, patterns start to emerge that show how humans drive in a particular area in a certain car in specific weather conditions. This data can then inform your driverless vehicle how to conduct itself.

The image above illustrates the different speeds of cars when driving along a straight road before navigating a bend.

Not shown are the weather conditions, the type of vehicle being used, and the time of day when the speed data was recorded. Correlating this information provides a clear insight into driving patterns for this curve.

But even without such correlation it becomes apparent that the speed on the curve is lower than at the straight road sections, even though the speed limit is the same for both the curved and straight sections of the road.

What are speed profiles?

HERE has already derived speed profiles from analysis of probe data obtained over long periods of time. While we currently use this analysis to provide more accurate traffic forecasting, this data could also be leveraged for autonomous driving. These speed profiles are created based on a complex processing structure that involves four key factors:

1. Some people like to drive fast while others don’t. A family car might by default have a “normal” speed profile and follow it accordingly. A sports car, on the other hand, might have a faster speed profile, with more rapid acceleration and deceleration.
2. Weather conditions affect how fast people drive so we create different profiles for dry sunny weather and for rainstorms.
3. When a traffic light causes many vehicles to stop and many others to drive with a mid-speed value we use some maths to exclude irrelevant data.
4. Sometimes a vehicle in front limits another vehicle’s speed to the point they are driving well below road regulations; in this case a speed profile isn’t relevant.

And it is just this kind of adaptability that could make autonomous driving a bespoke, human experience that can adapt to the needs and wants of its passengers. And if you’re in search of some jerky thrills, go and let off steam on the dodgems.

What do you think your speed profile looks like? And would you feel more comfortable in a car that doesn’t behave robotically?