Quote
nicholas.seward
@Hazer: Real quick. The prepocessor I have made breaks movement in 1mm or smaller chunks. (Actually whatever size you want but 1mm doesn't show any artifacts.) At that scale a straight line in one coordinate system is approximately straight in the other.
All the math is the same between the Rostock and LISA (LISA enjoys the same simplifications that the Rostock does.) except for the arccos you need to do for each arm. It seems like there are libraries that use lookup tables that can get that done in pretty short order. 17 cycles for tan. I am not sure how many for arccos/arcsin. [forums.reprap.org] I have no experience myself but I feel confident that if I wanted to squeeze LISA into AVR firmware that I could. (Worse case scenario is that I have to make my own lookup table.) However, I am not going to worry about it because it seems that the next gen controllers won't even flinch with unoptimized inverse kinematics.
Hi
There are cordic methods for all of the trig functions. They are reasonably compact and fairly fast. As long as you are not asking about something that goes to infinity, they work pretty well. Put another way - sin. cosine and tangent are fine, ATAN can have it's issues.
Bob