As part of the investigation in the feasibility of the hexapod project, I've spent the last month playing with a small hydraulics test system. The final system will need 18 actuators, all moving simultaneously and quickly, controlled fairly precisely. Is this possible? (Or, more the point, affordable on my budget?) It's an open question.
The first step was the hook up a small electric motor to a small cylinder and control if from an Arduino. This was done with a solenoid valve interfaced through a basic solenoid driver circuit.
Control of the valve is only one side of the equation, however. To move the joint to a specific position, you need to know where it is, so you know when to stop moving it! This requires an encoder. Out of curiosity I tried using a flex sensor -- I don't recommend it. They're super noisy. Much more promising was the AS5040 chip, which is a non-contact magnetic system. A friend had a test board which I could borrow. It worked much, much, much better. With the addition of a PID library, it was all more or less working.
So, I'm sure I can control the hydraulics as needed. At a much higher cost than I would like -- the valves are going to be at least $100 each! The next step is to rigorously spec out the entire system using real components. Specifically, the horizontal hip joint needs to be able to cycle fast enough, using a real cylinder, driven by a hydraulic pump supplying a realistic volume. The big question is determining just how much force will be needed there. I'll be prototyping the joint and performing empirical measurements with full loading over the next few weeks. And once I have that answer, I'll have a pretty solid guess of just how much this is all going to cost -- and thus if I can do it this year or if it will have to stretch into 2012.