MEAM Seminar: “Hierarchical Task-Parameterized Learning from Demonstration “

Many modern humanoid robots are designed to operate in human environments, like homes and hospitals. Such robots could help humans accomplish tasks and lower their physical and/or mental workload. However, robot users in homes and hospitals typically are not familiar with robotics or programming, therefore it is difficult for them to adapt robots to their specific needs and environments. To remedy this situation, many researchers turn to learning from demonstration (LfD), which enables a robot to emulate natural human movement as opposed to having an operator devise control policies and reprogram the robot for every new situation it encounters.

We suggest a hierarchical LfD structure of task-parameterized models, particularly for object movement tasks that are ubiquitous in everyday life and could benefit from robotic support. Inspired by the task-parameterized Gaussian mixture model (TP-GMM) algorithm, we develop the hierarchical structure and explicitly utilize task parameters to maximize the expected performance in a new situation from a few demonstrated situations. The robot can thus determine when it should request new demonstrations when the expected performance is too low. Other advantages of our approach include that a wider range of task situations can be modeled in the same framework without deteriorating performance and that adding or removing demonstrations incurs low computational load, and thus the robot’s skill library can be built incrementally. We show these advantages in a simulated task and in the real world where naïve participants collaborated with a Willow Garage PR2 robot to move a handheld object. For most tested scenarios our hierarchical method achieved significantly better task performance and subjective ratings than both a passive model with only gravity compensation and a single TP-GMM encoding all demonstrations.