What would be the one thing I would want to say about robotics? I guess there's a series of things. That's where I get into trouble. I can't narrow it down. I suppose I should start with artificial intelligence. An artificially intelligent machine has to be evolved. It can't be designed in one fell swoop. What is needed is a platform that can be steadily improved upon. It has to be the ultimate in simplicity. Probably the most important feature would be feedback. Feedback is the cornerstone of intelligence if intelligence is the ability to adapt. Of the five human senses that make feedback possibe the sense of touch is the most basic. You can be blind, deaf and dumb, but if you can't feel you're a vegetable. If you have motion and feedback from that motion then you're a player. You can interact with the world. You can start building the foundations of intelligence. Next you need an application. You have to have a purpose. Information is just applied data. Intelligence is applied information. Wisdom is applied intelligence etc. The most basic application would be reproduction. The machine has to be simple enough that it can participate in it's own reproduction yet complex enough that it can do useable work. Also it has to be cost effective enough that it can flourish. Once it can participate and in some sense control its reproduction then it can branch out to other uses and become more viable. It will eventually need mobility but in its nuturing stage it can start as stationary as long as it can reach out. To me, what it has to be able to do is grasp. In today's robotics end effectors are all designed for specific uses. The hand is considered superflous. I believe it is the cornerstone to intelligence. Robonaut uses a hand, but it is strictly remote control, and the space program has slowed to a crawl. We'll be waiting a long time for it to evolve. Asimov too has hands, but does not accentuate feedback. The beauty of Asimov is its motors. For an intelligent machine to evolve and flourish it has to as motor independent as possible. An intelligent machine will eventually be able to make something as complex as a motor, but that complexity has to be put off as long as possible.
The basic physical axiom of robotics is the strength to wieght ratio. If it not greater than one the robot cannot even move itself. If it is to do useable work it has to be much greater. The trick as I see it is to have the power source removed from the remote manipulator. This would indicate some sort of tendon control.
There is much more to be written down, but this should constitute a beginning.