A Modified Cartesian Space DMPs Model for Robot Motion Generation.

DMPs (dynamic movement primitives) are a method to generate trajectory planning or control signal for complex robot movements. Each DMP is a nonlinear dynamical system which can be used as a primitive action for complex movements. The origin DMPs are used to model the robot joint space motion, however in many cases, robot motions are defined in Cartesian space, the model of Cartesian space is necessary. A Cartesian space DMPs variant is proposed which adds a dynamical quaternions goal subsystem to make the generated cartesian space twist more smooth and steady in the initial stage in this paper. This DMPs variant can be useful in some robot tasks which often require low speed operations, such as contact operation.