Tesla CEO Elon Musk has admitted that the company has got a lot of hard work to get through if it's going to launch a robot taxi network in 2020.
In a tweet on Sunday, Musk said "massive effort" is required to get the technology to "99.9999% safety." He also highlighted two areas where the software needs to improve.
Read more: Tesla is taking direct aim at Uber and Lyft with claims it plans to roll out 1 million robo-taxis by next year — but the plan was lacking details
"Intersections with complex traffic lights & shopping mall parking lots are the two biggest software challenges," Musk explained.
The tech billionaire set out plans in April to "have over 1 million robo-taxis on the road" by 2020. In his tweets on Sunday, he was more specific, saying it would be "by end of" next year.
Tesla will want to finesse the technology after its Autopilot feature has been controversial for the company. Over the past few years, at least four people have been killed while using this system, raising questions over whether the technology is safe.
Despite this, Tesla maintains that vehicles with Autopilot engaged have lower accident rates than those that do not. "Our data shows that, when used properly by an attentive driver who is prepared to take control at all times, drivers supported by Autopilot are safer than those operating without assistance," a Tesla representative said in a recent statement to the press.
Sbatio on July 8th, 2019 at 16:27 UTC »
Let’s start with highways being automated driving and have drivers do the “last mile” driving.
Stuntz-X on July 8th, 2019 at 15:37 UTC »
Uh forget the shopping mall parking lot. That is like the king of road hazards all lumped into one place.
TheRealFalconFlurry on July 8th, 2019 at 14:25 UTC »
That thumbnail tho...
all joking aside, autonomous vehicles will be drastically limited in reliability and efficiency for as long as they have to share the road with human drivers and function in a system built for humans. The way machines handle transportation is fundamentally different from the way humans do, and the reason this technology so hard to develop is because we're essentially trying to retrofit it to a system it's ill suited for.
If the transportation system was built from the ground up with autonomous vehicles in mind the technology would not need to be nearly as complex, especially if it doesn't have to deal with human error.
intersections would not need traffic control; roads could be embedded with passive "beacons" to allow vehicles to receive location data and information about upcoming road conditions instead of relying on visual interpretations which is very cpu-intensive. To clarify, I mean the road could have these "beacons" embedded every X distance which when pinged by the car, the car would be able to read a set string of data which would tell it, for example: that there is a (specific type of) intersection in 300m; as well by triangulation the ping would tell the car its relative position on the road. This would require far less computations per second than the alternative of processing and interpreting visual data because it is digital data being read by a digital machine rather than analog data converted to digital.
Roads wouldn't need painted lanes and parking lots could assign available stalls to cars. That would alleviate much of the complexities of finding parking stalls.
Sadly this approach is quite a drastic one and will likely not occur at any time in the near future for many reasons.