With so many autonomy-enabling technologies and loungy car cockpits-of-the-future littering the CES show floor, and with so many Aptiv-branded Lyft BMWs shuttling attendees around Las Vegas autonomously, it’s easy to imagine that full level-5 commuting nirvana is mere months, not years away. Against this backdrop Gill Pratt, CEO of the Toyota Research Institute, tapped the brakes. In his refreshingly realistic assessment of the landscape, he explained the three programming challenges of autonomy:
- Perceive the environment
- Predict what’s about to happen
- Plan a safe path through it
Pratt believes the industry is getting pretty close to solving number 1, with loads of innovations on display at CES. Number 3 is a cinch. It’s that middle step where state-of-the-art artificial intelligence is still nowhere near matching the capabilities of a focused human brain under all conditions all the time, and getting to that point could take many years. So sure, we’ll see autonomous ride-hailing and package delivery services launch in meticulously mapped urban areas (with decent weather) in the very near future. Such commercial vehicles, in operation around the clock, can pay back the high investment cost of the elaborate sensor arrays required for true autonomy. Pratt also cautions against expecting the rollout of these AVs to dent our traffic fatality rate much.
But maybe the sensors and intelligence we have today can dent our 100-traffic-deaths-per-day rate during these interim years while we retain control of our cars. In other words, in an effort to not let the perfect be the enemy of the good, Pratt and TRI propose rolling out Guardian very soon.
Guardian works like a vigilant, attentive copilot or driving instructor, continuously checking your work as you drive, “perceiving” the environment via cameras, lidar, and radar, and monitoring your own attentiveness by watching your head and eye movements. Guardian will warn you of impending dangers, and then if you don’t react quickly or thoroughly enough, it will “plan” your optimal path through the danger and help you execute it by adding steering, brake, or accelerator inputs. It remains completely up to you to “predict” what your fellow flawed humans are going to do (and whether you’ll feel worse if you hit those nuns or that baby carriage).