Intel's self-driving car comes to a stop while telling me that a pedestrian is about to cross the street. I watch from the back seat as, sure enough, a guy steps off the curb and crosses in front of the car.
The point of this exercise wasn't to let Intel's engineers refine how the car responds to pedestrians, it was set up to gauge my reaction to what the car was doing.
As part of its efforts to build an end-to-end self-driving car system, Intel wants to humanize the experience for passengers. To that end, it conducted a small study of 10 to 11 participants at its Arizona campus near Phoenix. The participants summoned the car with an app then rode along a set route on the campus. The drive included various incidents set up by Intel, such as the pedestrian stop.
Instead of making riders feel completely at the mercy of the car, Intel implemented displays and audio alerts so the car could communicate. When the car showed up, a personalized message showed my name on a passenger window-mounted display so I wouldn't confuse it with other self-driving cars that might be picking up other passengers. Think of the crowds of taxis, Ubers and Lyfts that show up whenever a crowd leaves a large event.
The car showed its route to my destination on a screen set up for rear-seat passengers, reassuring me that it wouldn't be kidnapping me to some bad part of town. When it encountered the pedestrian, a voice alert explained why it was stopping. Later in the drive, a detour forced it to deviate from its original route, prompting a visual and auditory explanation of what was going on.
Intel recruited its participants from the general public and is recording their reactions to the car. The staff conducting the study will examine the recordings to determine whether the car gave the passengers a positive experience.
For Intel, a positive results means the participants felt they could trust the car. Matt Yurdana, creative director in the Internet of Things Experiences Group at Intel, says that this type of trust is made up of four feelings: Safety, Comfort, Confidence and "In-control." Passengers should feel safe in the car. They should have a comfortable experience, as opposed to one that induces stress, have confidence in the car's behavior, and also feel like they're in control of the experience.
Being in control when riding in a self-driving car may seem like an oxymoron, but Intel expects the car to respond to the passengers' requests. It should take passengers where they want to go, whether that means going to the initially requested destination or accommodating a sudden change of plans.
The cars used for the study only communicated with passengers through screens and audio alerts. But Jack Weast, Intel's chief architect for Autonomous Driving Solutions, envisions more two-way communication in the future. He says that Intel's self-driving cars could potentially respond to voice commands and gestures from passengers.
One thing that's easy to do with a human driver but currently difficult to contemplate with a self-driving car, for example, is fine-tuning the destination. Imagine if you told a self-driving car you wanted to go to Eastridge mall. It might take you to the parking lot entrance, at which point you might want to refine that destination, asking it to take you to the Macy's entrance.
Along with developing the typical components of a self-driving car, such as recognizing objects and pathfinding, Intel's study shows that its engineers are also thinking about the way these cars will interact with riders.
And that may be a crucial distinction. As companies such as Waymo, Apple, NXP and Continental, to name a few, develop self-driving car systems for automakers, Intel may have found the key to making its system feel less like Skynet, and more like R2-D2.