Policy Adaptation through Tactile Correction
نویسندگان
چکیده
Behavior adaptation based on execution experience can be a practical tool to increase the robustness of a robot behavior learned from demonstration. While demonstration learning is a powerful technique for the development of robot behaviors, in general development remains a challenge. This work presents an approach for policy improvement through a tactile interface located on the body of the robot. We introduce the Tactile Policy Correction (TPC) algorithm, that employs tactile feedback for the adaptation of a policy learned from demonstration. We provide an initial validation of refinement under the TPC algorithm on humanoid robot performing a grasp positioning task, and policy performance is found to improve with tactile corrections. We additionally show different modalities, namely teleoperation and tactile corrections, to provide information about allowable variability in the target behavior in different areas of the state space.
منابع مشابه
Tactile Guidance for Policy Adaptation
Demonstration learning is a powerful and practical technique to develop robot behaviors. Even so, development remains a challenge and possible demonstration limitations, for example correspondence issues between the robot and demonstrator, can degrade policy performance. This work presents an approach for policy improvement through a tactile interface located on the body of the robot. We introd...
متن کاملLearning from Demonstration and Correction via Multiple Modalities for a Humanoid Robot
This paper reports ongoing work that employs multiple demonstration modalities in order to accomplish motion control learning in a multi-staged policy adaptation process. A novel interface for providing tactile guidance to correct learned motion control behaviors is introduced. This interface extends our prior work by making use of a more sophisticated set of tactile sensors, developed by the R...
متن کاملDemonstration, Tactile Correction and Multiple Training Data Sources for Robot Motion Control
This work considers our approach to robot motion control learning from the standpoint of multiple data sources. Our paradigm derives data from human teachers providing task demonstrations and tactile corrections for policy refinement and reuse. We contribute a novel formalization for this data, and identify future directions for the algorithm to reason explicitly about differences in data source.
متن کاملInvestigating Visual, Proprioceptive and Tactile Interaction through Sensorimotor Adaptation
متن کامل
Design of Vibration Alert Interface Based on Tactile Adaptation Model to Vibration Stimulation
In this paper, we describe the influence of vibratory adaptation on vibration perception and its correction method. Vibration strength after change depends on the vibration frequency before change even if the vibration frequency itself does not change. We found that the higher the vibration frequency before change, the weaker the perceived vibration strength. To correct this perception gap, the...
متن کامل