A Dataset of DMPs for robot motion planning

Collaborative robots, working autonomously or in cooperation with humans, are used in industrial scenarios to help workers during the execution of repetitive and tiring activities that may lead to the onset of musculoskeletal disorders. This paper introduces a publicly available dataset of distinctive motion parameters (i.e. Dynamic Movement Primitives, DMPs) to be used for planning the motion of cooperative robots. Learning by Demonstration methods were adopted to compute DMP parameters for common working activities, i.e. handling good, hammering and screwing, executed by human demonstrators. The computed DMP parameters were then stored inside the dataset. Once known the task to be performed and the target to be reached, the most appropriate set of DMP parameters is extracted from the dataset and the numerical integration of the DMP equation returns the robot desired Cartesian reference. The dataset-based motion planner was tested in simulation: for each sub-task, several target points were used to plan trajectories. The obtained results, in terms of generalization capability of the motion planner, confirmed that the proposed approach is able to plan Cartesian trajectory for working activities with a mean Success Rate of 85% for the handling good, 67% for the hammering and 61% for the screwing tasks.