>>12335220Quadruped robots may not seem very useful today, but understanding the control of legged systems is crucial for robotics to spread into areas that require more flexibility than what wheeled locomotion can provide. Most experts of the field would name search & rescue tasks as their aim (e. g. The DARPA Robotics Challenge for autonomous robots was started after the Fukushima disaster with this in mind), which may seem quite niche, I agree, but evetually this will carry over into contsruction, agriculture and services as well.
There is a reason that research teams are starting spinoffs like Boston Dynamics (since 1992!) at MIT or ANYBotics at ETH Zürich; once these systems are capable of going mainstream, the few companies who are able to make them will make bank. Just like a lot of R&D stuff, this may take longer than anticipated, I'd wager this is why Google sold Boston Dynamics to Softbank, but idk.
>>Needs a human pilotThe "human pilot" is doing less than you'd imagine; he's just guiding the robot in general directions (maybe selecting gait patterns too). This means that given a well-defined task (explore a building site, drag object from point A to point B) should be relatively easy to program, and the reason why you must have an operator in there in showoffs/conferences is that these environments are filled with sneaky unpredictable humans.
Last week, I've just been to a live demo in one research centre focusing on legged robotics, and in the lab, they let the robots move autonomously - the researchers just send them to places in the lab and they walk there unguided.
If anyone is interested, I'm sharing the case study that we were shown in a class related to the subject. This might clear up things and help in understanding the challenges of the field.
https://www.youtube.com/watch?v=RN888X4serk&list=PLE-BQwvVGf8EUrca1RmKIjA9OGRJqaGad&index=33&ab_channel=RoboticSystemsLab 
