[JesseK]

I'll start this off by saying that I'm not an expert, but here's my experience with 'muscle wire', aka 'memory wire', aka Nitinol. These have deterred me from considering its use in FRC robots, though you may derive different conclusions should you try it yourself. I agree with Al's comments on heat dissipation; the wire itself loves to get toasty if you use it in heavy loading situations...much like a motor near stall. Altogether, here are the major drawbacks I found:

1.) Heat dissipation to air via direct convection was not so great. My application was underwater robot fin control, so putting the wire in a dissipative gel in a tube that would be constantly cooled by water solved that problem for me. For a FIRST FRC robot, you're not supposed to have exposed wire, and insulating memory wire to satisfy that requirement will definitely bring about the issue of heat.

2.) After 130 or so repetitions at ~33% of the max calculated 'load', the wire started fraying. After ~160 it snapped. The wire diameter was the equivalent of 24AWG wire, though it is rated for much less current than normal wire. Perhaps thicker or better-cooled wire has a longer life?

3.) The expansion and contraction of the wire was not immediate. Especially as the wire became more worn, the delay was more noticable. These delays are usually in graph form in the spec sheet of the wire.

Also, keep in mind for your application that when the wire expands, you can't expect your mechanism to automatically return to it's default position -- doing so is literally like pushing on a string. In other words, the nitinol wire returned to it's normal length, but doing so didn't exert any usable force. That said, any mechanism used to return your device to it's 'default' position (such as a spring) must always be taken into account when calculating loading during contraction.