Incremental local online Gaussian Mixture Regression for imitation learning of multiple tasks

Abstract

Abstract-Gaussian Mixture Regression has been shown to be a powerful and easy-to-tune regression technique for imitation learning of constrained motor tasks in robots. Yet, current formulations are not suited when one wants a robot to learn incrementally and online a variety of new contextdependant tasks whose number and complexity is not known at programming time, and when the demonstrator is not allowed to tell the system when he introduces a new task (but rather the system should infer this from the continuous sensorimotor context). In this paper, we show that this limitation can be addressed by introducing an Incremental, Local and Online variation of Gaussian Mixture Regression (ILO-GMR) which successfully allows a simulated robot to learn incrementally and online new motor tasks through modelling them locally as dynamical systems, and able to use the sensorimotor context to cope with the absence of categorical information both during demonstrations and when a reproduction is asked to the system. Moreover, we integrate a complementary statistical technique which allows the system to incrementally learn various tasks which can be intrinsically defined in different frames of reference, which we call framings, without the need to tell the system which particular framing should be used for each task: this is inferred automatically by the system.

Imitating operations on internal cognitive structures for language aquisition

“…ILO-GMR [18] would allow the immediate incorporation of new data (if a new demonstration is close to a group of movements it can be immediately added to that group without the need to re build a model as the models are built on line).…”

“…The paper [17] examines how a task can be divided into subtasks, in a way that is related to the problem of finding the number of movements. Previous work [18] examined Incremental Local Online GMR (ILO-GMR) that can learn an open ended number of tasks from unlabeled demonstrations. In [18], the 2D position of an object was used to determine what task should be performed.…”

“…Previous work [18] examined Incremental Local Online GMR (ILO-GMR) that can learn an open ended number of tasks from unlabeled demonstrations. In [18], the 2D position of an object was used to determine what task should be performed. This object position was used in a way which is similar to the communicative signs in the experiment presented here.…”

“…The average of the output of these points is used. More sophisticated methods could easily be inserted here, for example ILO-GMR [18] or GMR [12] together with BIC [15]. Since low dimensional and accurate data is available after a successful grouping algorithm more sophisticated methods are not needed in this case.…”

Online quantum mixture regression for trajectory learning by demonstration

“…Regarding the comparative merits of both methods, it has been shown that GMR performs better for low dimensional demonstrations [13], while LWPR should be preferred for inputs of high dimensionallity, which lie in lower dimensional manifolds, and/or inputs that may contain irrelevant dimensions.…”