In 2010, I was an intern at Google working on a team doing the very early prototypes of touch screen stuff for ChromeOS. I was in the room the first time we scrolled the screen with a finger (initially, the wrong way). That same week, a bunch of tech blogs with "insider information" announced we'd have a consumer ChromeOS tablet by the end of the month. Spoiler: they were 8 years ahead of themselves.
Not to mention, the auto is doing something useful simply by moving. Perhaps a security robot in the home could be valuable by simply moving, but otherwise, it needs an additional layer of application. Interacting with humans, manipulating objects which it would have to identify, ...
It will probably be a flat fee based on forcasted average uses. Light users will subsidize heavy users. Like all sorts of services that come before it. Amazon is already great at this.
It is marketing hell to introduce some type of credit per use system. Adds too much stress on end-users.
I think the project would be a blast to work on and I certainly think research in this area should be supported, but I would love to know what they want to accomplish that couldn't be done by selling 2 or 3x additional static devices for less.
I also would appreciate robots that empty the dishwasher; move the laundry from the washing machine to the dryer, and then put the dry items into closets and chests of drawers; wash windows; clean ovens and fridges; and so on. Basically furthering domestic robots that free us from routine labor, just as the current, limited robots have done for us (I tell my daughters the washing machines are simple robots, and we should expect better in the near future).
Amazon has 1 partially solved with kiva, 2 with Alexa, leaving the big problem as 3. Amazon is well aware of 3 though because that is their main problem blocking them from automating hundreds of thousands of jobs in their warehouses. They even hold competitions for it: https://www.wired.com/story/grasping-robots-compete-to-rule-...
With all of that potential labor savings in solving that problem, and the possibility of a home robot using the same technology, I doubt anyone spends more R&D on this problem.
And I would separate your #3 into two items: recognizing household objects (a very significant bit of work), and then interacting with individual objects. Single instances of each of those are tasks that whole companies are devoted to. General-case systems are just not in the cards at this point.
Robots currently avaiable in that space are highly customised to a specific task (eg: Roomba). My roomba has trouble with wires, long hair, rubber bands ect. I have to do a "prep" before I turn that thing on so it doesn't trip. Defeats the whole point of it, prep is worst part about vacuuming, I might just finish the whole thing myself by using the manual vaccum.
In my opinion this is a bit creepy. Now one would be putting an Amazon device capable of sensing and actuation in their home. And not just sensing, enough sensing to map a house and perhaps keep track of where people are. If the robot does image recognition, so one can say hold up an empty box to buy more on amazon, that's even creepier. As the only practical means of doing this would be to send images to amazon's servers.
SDC - Self Driving Cars
Really SDC is much more of a software thing and surprising we have not seen other big tech playing in the SDC space. Some rumors on Apple but not shown anything.
An Alexa-enabled roomba clearly sounds like the lesser evil. And it could actually sit quite nicely in a the previously unoccupied intersection between the Alexa and the Amazon Basics subbrands.