The recent integration of the Make controller with Flash, its scripting language and image processing capabilities, solves many of the problems I was dealing with, from the choice of microcontroller to the controlling software.
The program currently reads the position of the performer on stage and even the distance from head to ground, all with only one cheap webcam as an input device. There is so much information that can be acquired through image analysis that effectively eliminates the need for a variety of sensors that I had planned to use.
The only limitation is the amount of servo ports in the controller. I will have to daisy-chain at least two, to handle all the servos I think I need.
The servo that supports the weight of the marionette most of the time needs much higher torque than the ones I have. I have not calculated the minimum torque yet, but need to do that ASAP to order it and start testing the software that Philip and I are designing. I am very excited and happy with the progress so far. At first I had my doubts that the solution he proposed would work, but now I see it can, and in a very simple and elegant way.
The two sizes of spools we currently use (5 and 8 cm diameter) happen to be just perfect to provide most of the travel that I need without using continuous rotation motors instead of standard servos. This simplifies programing. We found out by testing a variety of "identical" modified servo motors that the speed of rotation, given equal parameters, is quite different, which implies writing routines to address each one of them to synchronize, a royal pain!
The only continuous rotation motors I need are for the arms since they travel a much greater distance.