Automated Trajectory Morphing For Marionettes Using Trajectory Optimization

We present a method to automatically morph trajectories from one mechanical system into trajectories of another system with potentially very different dynamics. Our method relies on a mapping from the source system’s configuration space to that of the target system to create a desired trajectory for the target system. A projection operatorbased trajectory optimization finds a dynamically-admissible trajectory for the target system that approximates the desired trajectory. The optimization algorithm is extremely robust and often finds a satisfactory trajectory even for highly nonlinear systems with instabilities, uncontrollable modes, or closed kinematic loops. This robustness simplifies the mapping step by essentially allowing us to ignore the target’s dynamics and focus on the purpose of the trajectory. We demonstrate the process using a marionette as an example. The marionette has complex dynamics, uncontrollable modes, and closed kinematic chains. The trajectory morphing process allows us to specify trajectories for the puppet using human motion capture or animation tools without considering the actual abilities of the puppet.