Improving Task-Parameterised Movement Learning Generalisation with Frame-Weighted Trajectory Generation

Abstract

Learning from Demonstration depends on a robot learner generalising its learned model to unseen conditions, as it is not feasible for a person to provide a demonstration set that accounts for all possible variations in non-trivial tasks. While there are many learning methods that can handle interpolation of observed data effectively, extrapolation from observed data offers a much greater challenge. To address this problem of generalisation, this paper proposes a modified Task-Parameterised Gaussian Mixture Regression method that considers the relevance of task parameters during trajectory generation, as determined by variance in the data. The benefits of the proposed method are first explored using a simulated reaching task data set. Here it is shown that the proposed method offers far-reaching, low-error extrapolation abilities that are different in nature to existing learning methods. Data collected from novice users for a real-world manipulation task is then considered, where it is shown that the proposed method is able to effectively reduce grasping performance errors by ∼ 30% and extrapolate to unseen grasp targets under realworld conditions. These results indicate the proposed method serves to benefit novice users by placing less reliance on the user to provide high quality demonstration data sets.

A robot learning from demonstration framework for skillful small parts assembly

“…In order to further boost the generalization capability of the approaching policy P app , i.e. to preserve the local structure in demonstration around the approaching frame, the frame-weighted TP-GMR algorithm (αTP-GMR) [25] is utilized here instead of the typical one. In the j-th frame, the Gaussian center µ and covariace Σ can be partitioned as…”

Learning Task-Parameterized Skills from Few Demonstrations

“…By learning the forcing term in DMP with TP-GMM, [16] presents an approach that resolves the divergence problem usually associated with GMM type models. [17] introduces weighting to TP-GMM and [18] build on the idea and proposes to infer weights from co-variances in the normal distribution. While the focus of above-mentioned improvements is on better policy generalization instead of data efficiency, our paper alternatively addresses the latter which is an equally important problem in LfD.…”

Learning to Shift Attention for Motion Generation

“…However, the combination of Gaussian components from local frames distorts the generated global trajectories, especially when they are far away from the those in human demonstrations. In [14], [15], [16], the authors solve this problem by introducing a variance division factor to make the low variance component more valuable in Gaussian multiplication. The assumption is that the local frame with a lower variance is more important than the one with a higher variance.…”

“…The assumption is that the local frame with a lower variance is more important than the one with a higher variance. In [16], the variance division factor α for each frame is dependent on the variance of their local trajectories Σ Σ Σ. For the k-th local frame,…”

“…1) Variance Weighted Reproduction Cost: In [16], the author suggested a variance weighted reproduction cost to determine the parameter γ in Equation 1. We use a similar cost for the attention model.…”