> delivery robots are easier to accommodate to the point of actually becoming economical

Delivery robot co-founder here (Robby.io, YC S16). The economics are actually really, really hard.

1. A Honda Civic costs $24K. It's actually surprisingly difficult to build a sidewalk-sized robot for less than $15K if you include labor for assembly and if you don't have scale. And it isn't compute and sensors. It's everything else -- plastic, metal, structural stuff, crimping wires, motor mounts, gaskets, all of that. The form factor isn't something that is mass-produced already.

2. They can only deliver MAXIMUM 2km, anything beyond that the food gets cold. Even 2km is a stretch, customers complain of cold pizza. A lot of the existing human-driven delivery market comes from further deliveries, because you know, if you can walk to the restaurant you might actually be willing to take that walk.

3. Real estate for robot storage is a HUGE cost issue in the bay area. Rental warehouses are not cheap in Mountain View. And in places where rent is cheap, people tend to also be less busy and not need food delivered.

4. Teleoperation is also a huge cost. Full autonomy on sidewalks is not difficult, but dealing with human issues at stop signs and intersections is a very difficult issue to solve.

5. Getting restaurant staff to be willing to walk outside the restaurant and drop food into a robot is an issue. The current state of many restaurants is that they just put orders on a shelf and delivery drivers walk in and pick them up without intervention. They just don't have the staff to deal with a robot that can't come inside. Requiring staff to find the food, walk outside, find the right robot, and drop it in would be a regression from their point of view.

All this stuff adds up, quickly.

I see basically zero chance of consumer delivery robots being economical. Self-driving cars also have a low chance of panning out IMO -- it will not take many more incidents to kill the industry -- but if self-driving cars did pan out it would at least be a service I would use.

In contrast to SDCs, the delivery robots I have seen are bulky and take up a lot of sidewalk space. In contrast to delivery people, they don't climb stairs. If they did climb stairs, I don't want to be stuck behind one. As with SDCs, most people hate these things.

I think if they were humanoid-shaped it would help with some issues (take less space, climb stairs) but introduce other issues (less carrying capacity, far more complex/expensive). The only consumer-facing robot I've seen so far that solves a problem in my life, is the Roomba+competitors. I'm pretty skeptical that consumer robotics will be a big part of the future.

> Compared to self-driving cars, delivery robots are easier to accommodate to the point of actually becoming economical.

Seeing how delivery companies abuse labor laws and take advantage of undocumented immigrants to perform their service, I'd say delivery robots are a far more ethical solution as well.

This is basically not true. Low end automobiles and solutions like ebikes, etc have the volume to sustain production at crazy low margins. Despite tech fantasies, delivery bots, drones, etc. aren't even close.