Hardly a week passes without some announcement from automakers, suppliers -- even ride-hailing services -- about the impending arrival of self-driving cars.
But experts say a fully automated vehicle that is 100 percent safe 100 percent of the time and can operate on any street in any weather condition in the U.S. is not right around the corner.
It's a decade or more down the road.
That assessment, from Raj Rajkumar, co-director of the General Motors-Carnegie Mellon Autonomous Driving Collaborative Research Lab, is shared by other experts at other schools and elsewhere. Carnegie Mellon, the crucible of autonomous vehicle technology, has been working on self-driving vehicles since the 1980s.
The obstacles to perfecting and mass producing fully automated vehicles that can safely transport a passenger door-to-door with no human intervention are formidable: Sensing equipment, such as cameras, lidar and radar, has to get more efficient, especially in inclement weather. It also must get less expensive.
- Software has to be perfected that links the vehicle's controls with all the sensing hardware. And this software must be able to anticipate nearly every scenario a vehicle can encounter, from inclement weather to a traffic cop's hand signals to a pedestrian darting into traffic.
- Infrastructure needs to be improved, from lane markings to traffic signals to bridges -- as well as the vehicle-to-vehicle and vehicle-to-infrastructure communication systems. For a vehicle to drive itself safely in all conditions and speeds, it has to know where it is at all times so that, for example, it anticipates a stop sign around a corner.