fatality. broke the rules. NOTABUG.
We have radar and lidar, tried and tested technologies being used for decades. Unlike the wetware you speak of, this system doesn't get tired or distracted.
Yes motorcycles are smaller than cars, but that won't be a problem for radar/lidar, and with proper sensor placement, there won't be any blind spots.
Not sure how you can say that, given recent accidents (crash into pedestrian pushing a bike across a street, crash into a stationary lane divider). Possibly other challenges are even harder, but describing a task that real world driverless cars repeatedly fail in as "the easiest" strikes me as overly cheery/Pollyannaish.
It boils down to execution and application of the technology. The key elements, radar and lidar make it really easy to detect objects around the vehicle in real time and react quickly. Now how an autonomous vehicle company decides to use these technologies matters in the overall effectiveness. For example, Tesla doesn't use the lidar at all for cost savings. Tesla relies on radar and machine vision for localization which to someone like me is scary. On the other hand, waymo uses lidar. Look at their accident rates, very impressive indeed.
1. Uber crash, Tempe, Arizona, 2018-03-18
2. Tesla crash, Mountain View, 2018-03-23
> Autopilot may not detect stationary vehicles at highway speeds and it cannot detect some objects.
To be more precise, then: I agree with you that might be "very easy" to detect objects and then, say, stop. But apparently, self-driving cars would then stop all the time needlessly, which is why they're not programmed that way.
So, what is hard, then, it appears, is to avoid crashing into people or objects, but still keep driving when safe.
[1] https://en.wikipedia.org/wiki/Death_of_Elaine_Herzberg
[2] https://en.wikipedia.org/wiki/Tesla_Autopilot#Mountain_View,...
I expect a pivot (>10y) to wetware if people really do* demand to give up their dv/dt. Rat neurons were flying F-22 sims a decade ago.
*(bad idea, I don't think we will)