Astro robot (sort of available from Amazon.com; if they invite you to purchase) has no arm or ability to manipulate things. The robot can follow voice commands, avoid obstacles and find people very well.
Snacky is an accessory, allowing the robot to fetch snacks on command. The yellow (3d printed) piece in the cargo area tells Snacky to drop a snack (long end up) or not drop a snack (short end up). I received Astro on Saturday and completed the design and construction of Snacky by Wednesday--it's not too complex or costly (if you can access a 3d printer).
The setup is fairly simple:
For construction details, go here.
I received Astro on Saturday and completed the snack dispenser on Wednesday, using available parts. The IR detectors were not my first choice, but I didn't have ultrasonic sensors on hand. The IR sensors go crazy when Astro approaches--something being emitted from the front of the robot gets their attention.
Astro can carry about 4 pounds and has a usb port available for power. Notable abilities include mapping rooms, navigating around changing obstacles, and finding specific people within the rooms. I see significant potential involving auxiliary devices that load/unload Astro's cargo bin while moving things from fixed locations to/from random locations of humans.
I suspect that the first "home robots" will be like Astro; wheeled mobile devices that can operate with auxiliary units (like drink dispensers, snack dispensers, trash cans, tool delivery, etc.). Humanoid type machines such as this (picture below) will cost from $45,000. to $175,000., have minimal autonomous capability and not be available until 2024. Compare Astro (at $1,000. to $1,500) with a humanoid (at $110,000. to $175,000.) -- not inexpensive, but reasonable for the capabilities and build quality.
The humanoid machines will be appropriate (maybe even cost effective) for some industrial situations.