The most successful robot in the world is probably a washing machine or dishwasher, not an iRobot.

If you mean "will be" rather than "is", I'd agree. Otherwise, it's just a kind of reductio ad absurdum to stretch the definition of "robot" to include any machine.

The key differentiating factors (any one will do) seem to be resemablance to humans/animals (not all that relevant here, I'd expect), being self-propelled, and being able to perform complex tasks.

It's debatable how to define "complex", but I'd say that, for example, telling time, as a sibling comment suggests, does not qualify, no matter how complex the mechanism required to do it well happens to be.

Similarly, a machine that washes items that have been carefully loaded (and maybe even pre-treated by scraping food debris) by a human, such as a current dishwasher, is also not performing a complex task. Doing so to a haphazardly-filled bus tray or sink, however, would be.

The most successful robot in the world is the clock.

Robotics companies tend not to be unicorns because 1) they require monetary input to scale and 2) you can't sell them to a FAANG.

Both Amazon and Google have bought robotics companies.

FWIW google divested itself of most of its robot investments, though there are still people at X (at least) working on autonomous systems.

Knightscope is on its way to unicorndom, I believe, but overall I'd say robots tend to be fairly specialized and therefore each go after smaller TAM than most of unicorns out there today, many of which are agnostic to their customers industry.