Trajectory planning for robot arm based on the Improved Mixture of Motors Primitives

Abstract

In order to reduce the dependence of the dynamic motion primitive model’s trajectory planning on data sets and improve its generalization ability on small data sets, we propose an improved Mixture of Motors Primitives (MoMP) algorithm. MoMP uses a new motion primitive model to achieve the same direction as the taught trajectory and the learned trajectory by using less teaching information as components to build the original motion primitive information base. Additionally, MoMP uses the gating unit to develop an optimal weighting strategy to learn new motion primitives and form the motion trajectory. Using MATLAB software combined with the Robotics Toolbox to build a simulation platform, the iiwa robotic arm was utilized to plan the grasping path of a given object. As a result, the endpoint error in the planned motion was reduced by 48%.