[AutonomousVehicles] How Self Driving Cars will Talk to Pedestrians

[Cornelius Butler]

Hi Everyone,
Here is a great article on mashable about self driving cars needing to be
able to "talk" to pedestrians. Article link and text are below.


Article Link:
https://mashable.com/article/auotnomous-vehicles-talking-to-pedestrians

Article Text:

How self-driving cars will 'talk' to pedestrians
When there’s no driver you have to get creative with light, sounds, and
visuals.
By Sasha Lekach  on December 9, 2021


It's hard to communicate with a gesture or nod at a passing car when
there's no one at the wheel.

When it comes to driverless cars, this is a bigger, and potentially more
life-threatening, issue than it may seem. Road safety relies on all types
of communication between people and vehicles, and most importantly on a
psychologist-coined term, "communicating intent."

As pedestrians and fellow drivers we're constantly indicating what we're
about to do, whether that's with our words, body positions, or hand
gestures. Studies have shown that up to 84 percent of pedestrians attempt
to make eye contact with drivers before crossing in front of them. So now
robo-cars need to be programmed to show what they're planning to do — and
pedestrians will need alternatives to smiles, glances, and frantic hand
waves.

Self-driving cars also need to be able to communicate with paying
customers, otherwise it'll be a slow and messy process just matching
passengers with a ride.

Amazon-owned Zoox is well aware of these driverless limitations as it
builds a robo-taxi that will never have a human driver in the front seat.
There isn’t even a steering wheel. Instead of retraining the car to
function without a driver, Zoox is creating a new driving system from the
ground up.

The truly driverless vehicle allows Zoox’s lead product experience manager
Riccardo Giraldi to explore how future users and outside pedestrians,
cyclists, and other (human) drivers will communicate with each other.

“We’re making sure we make an experience that is mindful of everybody,”
Giraldi said about the development process.

With today’s human-driven vehicles, we use light, sound (think of all the
honking on the road), hand and head gestures, eye contact, and every road
rager’s favorite tool, voice, to communicate while in a car. Giraldi wants
to build “something better than what we are used to today,” but isn’t
completely unfamiliar.

He's experimenting with sound and light cues to notify those around the
vehicle what the Zoox vehicle is planning to do. He said the Zoox car is a
new entity on the road — it's not a person, bicycle, or the usual car or
even the usual self-driving car with a safety driver up front — so it will
need to gain everyone's trust. That starts with interacting and
communicating without forcing everyone to learn a new language just to
cross the street.

He talked about creating a "natural interaction" that's intuitive, "so you
don’t have to learn how to interact with it." So instead of learning a new
color-coded light system or figuring out where to even look on the car,
Zoox wants to subtly convey information through the blinkers or with a
simple sound. The colored lights on a Zoox car are similar to turn signals
and brake lights, but can produce a diverse range of pulses and color
combinations that, in theory, could communicate all sorts of things.

Other self-driving car companies are thinking more about the interactions
between robot cars and their potential customers. Waymo already has fully
driverless rides available in the Phoenix metro area. Earlier this year it
started using "Car ID" to give people a way to quickly spot the
self-driving taxi. Now a digital dash displays whatever the user has
programmed to light up.

I could type in my initials in the app and then that'd be displayed on the
car – or I could honk the vehicle's horn remotely through the app.

When it comes to interactions with passing pedestrians, a Waymo
spokesperson said safety and courtesy are being programmed into the Waymo
driving platform in the form of "subtle cues" rather than explicit
communication. For example, Waymo cars try to give cyclists a wider berth
when passing, or pedestrians more time and space at the crosswalk.

The makers of Cruise's self-driving Chevy Bolt EVs and future fully
driverless autonomous shuttle, the Origin, are also concerned about
interactions with pedestrians.

Brandon Basso, senior director of motion planning and control on Cruise's
robotics team, said in a recent phone call that Cruise’s focus is on that
same psychological term "communicating intent." He said, "The primary way
we signal to other road users is our actions."

So a Cruise Bolt might do something like moving over to give room to a
bicycle coming from behind. That doesn't require any extra signaling,
sound, or visual cues.

Cruise last month started testing fully driverless rides in San Francisco
with employees as the first riders (although not without some controversy
among local officials). As seen in some of the initial videos from the
rides, an interaction within the mobile phone app is Cruise’s solution to
the problem of robot-drivers not being able to spot their fares and let
them know they’ve arrived.

But for passers-by, Basso recommends treating self-driving taxis as
cautiously and warily as any other car on the road, especially since Cruise
studies real-life interactions between pedestrians, cyclists, and other
drivers to determine how it programs its cars to behave. There's no need to
act weird and hesitant around a self-driving car.

For other motorists, "bear in mind the Cruise vehicle is designed to comply
with the way in which the roadway is set up." That's Basso's way of
defending what human drivers might perceive as overly cautious and slow
driving.

It's also why Cruise has "self-driver in training” bumper stickers on all
its cars. The short phrase is the clearest way to communicate what it's
like to be around a driverless robo-taxi.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://nfbnet.org/pipermail/autonomousvehicles_nfbnet.org/attachments/20211213/88cdfd9b/attachment.html>