How do they react and get out of the way of an emergency vehicle?
In general though, I don't think this is something that automated vehicles will have trouble with. Picking out strobing lights is pretty trivial for computer vision, and a couple of external microphones located on opposite sides of the vehicle would have an easy time localizing the direction of sirens even before the EV is in line-of-sight. I suspect eventually there will be some sort of preemption signal that EVs will be able to transmit that communicates their intended route to vehicles in their path.
I look forward to the day that I have a big ol' "get out of my way" button on the dashboard of my ambulance...
I think I saw a light change on demand in my hometown (Wake County, NC) for an ambulance the other day. It was late at night with minimal traffic and the light had just turned red ahead of me. The ambulance was between me and the light and the light immediately cycled back to green as the ambulance approached. I'd never seen or heard of this before, but do you know whether emergency vehicles are now able to influence traffic signals?
edit: apparently so - https://en.m.wikipedia.org/wiki/Traffic_signal_preemption
One such article: https://boingboing.net/2006/04/18/man-fined-50-for-usi.html
By turning the light green, it allows cars that might otherwise be obstructing the intersection to proceed through it, clearing a path.
It's also pretty likely that whatever framework develops to license the systems will address liability (I think by requiring the operator of record to insure the vehicle).
All precedent is that auto manufacturers are responsible for manufacturing defects that lead to injury or death, not the owner or driver. Even the term "operator of record" has new meaning in the fully autonomous era. Bottom line if software bugs cause injury it's the software writers that are responsible.
You mention many problems with human drivers. But software has been known to have bugs too, and the world is an almost infinitely complex learning environment. It is most likely that the safer future will involve the intersection and mutual augmentation of these two control systems, not the swift replacement of one by the other. As is the case for example with aircraft.
How often is the driver in those cases under insured? I also wonder how often criminal charges pop up for drivers that weren't being negligent.
As for criminal charges, the key question is whether a bad bug can constitute negligence. That will no doubt be tested in courts when lives have been lost.
What does 'fully insured' mean here? A sensible approach is to have liability coverage sufficient to protect your assets. That might be a lot more coverage than the legal minimum, but a large personal umbrella policy might be hundreds of dollars a year. People go bankrupt because they skimp on insurance, not because it is onerously expensive.
Not necessarily. Even "large" policies will not protect your assets from the worst cases where damages run into the millions of dollars, e.g. where you kill a family or hospitalize someone for months.
You propose that everyone with coverage less than "sufficient to protect your assets" is guilty of being "under insured". But there is no amount of coverage you can pre-determine to protect your assets in all possible cases. Not unless you purchase a policy with no upper limit -- but those are rare in the auto insurance world.
Thus, even if you have a "large" policy that falls on the high end of typical coverage limits, truly catastrophic cases often result in personal bankruptcies. Bankruptcy courts cap the payouts. There are no such caps on auto manufacturers responsible for accidents... unless they go bankrupt too.
I would expect many of the largest settlements to involve negligence like drunk driving or excessive speed (a system might have the wrong speed for a stretch of road, but it mostly will have accurate speeds). Of course there are also ways that a driving system could be implemented negligently that don't apply to a human driver (but I sort of expect the licensing process to address that).
You need only look up the average cost of a major hospitalization to gauge what the at-fault party is responsible for. They can run into the millions.[1] That's before we even get to tort damages awarded by a court for things like negligence (and you're certainly right that those can be even higher).
[1] http://www.rslfunding.com/whats-the-average-car-accident-set...
Also, I'm sure there are large claims. My contention is that there aren't a lot of them.
As for liability, there is already generally strict liability when products malfunction. If the product didn't malfunction, then that would mean someone else was at fault, and the cases where there was no malfunction and no one else at fault should be rare. Regardless, there's a huge body of law out there regarding product liability.
Once again, I'm assuming that autonomous vehicles deliver on the safety promises. If they become as safe as a lot of people think they can be, there will be so many fewer accidents that I don't think insurance and liability will be as big of an issue as they are now for human driven vehicles.
The answer is: it's complicated. That's why we have courts. We don't even have it sorted out for human drivers yet; if you kill someone with your car, you may get off without even a ticket, you may end up in prison for manslaughter. I honestly don't think robotic drivers add any complexity: there's already infinite complexity ;)
It's going to be some time before both of them are reality.
Also I don't think police are going to be too happy about a vehicle that is going to be probabilistic about pulling over. Nor would I really be happy about that in the back of an ambulance.