Task Extension of Robot with Voice Control Based on Dynamical Movement Primitives

Since traditional dynamic movement primitive has good generalization ability, it is often used to generalize trajectories. However, it only applies to generalizing a simple trajectory which is used for a simple task. If we need the robot to accomplish other task, we have to reteach the robot a trajectory, which will consume time and manpower. In addition, the teaching trajectories of general tasks are often complex, and there are uncertainties in task constraints, so it is difficult to directly model and generalize the whole trajectory. In this case, the complex skills (trajectories) can be divided into a series of relatively simple sub-skills, and then the corresponding sub-skills can be modeled according to specific task requirements. In this paper, a task is taught to the robot, and the sub-trajectories obtained by task trajectory segmentation are generalized by DMP and arranged and combined in different order so as to expand to another task, greatly expanding the generalization ability of DMP. In addition, we combined voice control into the operation, so that we can control the robot by voice command to accomplish the task we want. Baxter robot was used to carry out experiments, and the proposed method is proved to be effective which can be seen from the experimental results.