Learning By Demonstration

The remote teleoperation of robots is one of the dominant modes of robot control in applications involving hazardous environments, including space. Here, a user is equipped with an interface that conveys the sensory information being collected by the robot and allows the user to command the robot's actions. The difficulty with this form of interface is the degree of fatigue that is experienced by the user, often within a short period of time. To alleviate this problem, we are working with our colleagues at the NASA Johnson Space Center to develop user interfaces that anticipate the actions of the user, allowing the robot to aid in the partial performance of the task, or even to learn how to perform entire tasks autonomously.

Our approach is to use our automatic control techniques to aid in the recognition of the user's intentions. Prior to the user demonstration, the control system enumerates the different grasping actions that can be used for each object in the workspace (essentially, the robot "imagines" what it would feel like to pick up every object). The movements produced by the user are then compared against each of these imagined actions. The one action that best matches the user-driven movement is considered to be the explanation of that movement. Using this technique, we are able to recognize entire sequences of actions.

This movie shows an example of a behavioral sequence demonstrated by a user through a teleoperation interface. In this sequence, the extracted sequence is: pick up the blue ball; place it on the pink target, pick up the yellow ball, and place it on the orange target.

This movie shows the automated replay of the same action sequence in a novel situation. Note that the movements are smoother and are executed more quickly than when the user is in control.

---

Interpreting Teleoperator Intent

During teleoperation the robot can monitor the actions taken, parsing sequences into behavioral schema.

The robot can use the apparent control references of these actions to discern the behavioral goals---the intention of the teleoperator.

The robot can then playback this behavior, utilizing its own prior, common-sense, knowledge about the task. The resulting actions are not a rote replay of the demonstrated sequence.

In this example, a sorting task is demonstrated by the teleoperator. Orange balls go on the right, blue balls on the left. The robot matches the intentions of the human by associating the human's goals. Notice how the human only demonstrates the task using only one arm.

Here the robot uses its own internal knowledge about how to move objects around to play-back the demonstrationg to sort the balls.