Super interessantes Video, wie sich ein autonomes Fahrzeug im Straßendschungel von Las Vegas mit Fußgängern schlägt.
As is usually the case with autonomous vehicles, it all starts with the sensors. Torc Robotics’ prototypes use the same trifecta of cameras, radar, and lidar as most other self-driving cars. The radar is also equipped with NXP TEF810x and MPC577xK chips that Torc claims “boost its perception capabilities.” This makes it easier to identify a person, but the car also has to know what to do once that happens.
Torc noted that pedestrians can’t always be counted on to follow the rules, so once one of its autonomous cars identifies one, it takes a more cautious approach until it knows what the person is doing.
“We balance assertiveness and safety around other vehicles … with caution around pedestrians,” Torc says in the video.
A fatal March crash involving an Uber self-driving car demonstrated what can go wrong when autonomous vehicles and pedestrians share the road. Traveling down a street in Tempe, Arizona, at night, the Uber vehicle struck and killed pedestrian Elaine Herzberg, who was pushing a bicycle across the street outside of a crosswalk. Neither the car nor the onboard human backup driver reacted in time to prevent the collision.
Whether self-driving cars can reliably avoid collisions with pedestrians is an open question, as is how cars with no human driver behind the wheel will interact with the people they share streets with. With a person in the car to, for example, wave to indicate they are yielding the right of way, things might get confusing. Ford has gone as far as dressing up a person as a car seat to test how this will all work.