My dad used to tell me "It's a lot harder to carpet the world than it is to wear shoes."

Ambitious redesigns of existing infrastructure are neat, but they are rarely more efficient or practical. Especially when you are overengineering to solve an issue that's already been dealt with. A self cleaning room requires a lot of additional hardware, all of which has to be designed, built and installed, and has to be powered and run by software that needs to be programmed. It also needs to be maintained, and depending on how it's cleaning things, it may also be dangerous, or at least capable of damaging property (ever have a motion activated light turnoff while in a bathroom stall? now imagine it triggers steam jets). Not to mention the potential hazards of water damage on a room if anything goes wrong.

Or, you can buy a mop for 0.1% of the price.

Humanoid robots can escape this problem because versatility adds value. The upfront cost may be tens of thousands of dollars, but for that price you're getting something that solves many, many problems. They can potentially go from task to task, filling a multitude of roles, and ideally with minimal down time.

It also helps that we can use existing processes to train them. They can observe human workers performing a task, attempt to replicate that task, and use feedback to improve. And that's critical because the hardware is the easier part, it's software that's the real challenge.