Motion Programs for Puppet Choreography and Control.

This paper presents a motion description language (MDLp) for specifying and encoding autonomous puppetry plays in a manner that is faithful to the way puppetry choreography is currently formulated. In particular, MDLp is a formal language whose strings, when parsed by a dynamical system (the puppet) produces optimized, hybrid control laws corresponding to strings of motions, locations, and temporal durations for each agent. The paper is concerned with the development of this language as well as with an optimization engine for hybrid optimal control of MDLp strings, and with the generation of motion primitives within the "Imitate, Simplify, Exaggerate" puppetry paradigm. [ABSTRACT FROM AUTHOR]