The other big hurdle is power. I've read that ASIMO's battery only lasts 30 minutes to an hour, which is far from ideal. Hopefully some of the recent battery developments we've been hearing so much about will make it to market in some usable form.
I think that's a likely approach. Also, I have the feeling that many of the first robots are going to work as multiples in unison. That could also address the battery issue too, they could take turns recharging. I'm trying to think what the first civilian applications might be. Probably a novelty restaurant where the wait staff are robots. :)
Another interesting thought: maybe there will be a market for different AI services where you pay for a connection to an offsite supercomputer to handle the robot's heavier processes. Each of these services will compete for the highest IQ score. I can see the advertisements now: Apple's patented Siri v9 is the only AI that has passed both the Turing test /and/ the Voight-Kampff test. For only $199 a month, why not add a Siri of your own to the family? [cue the delightful music and shots of friendly robots playing with children]