Contextual Task Recognition to Assist Mobile Robot Teleoperation with Introspective Estimation Using Gaussian Process

摘要

To appropriately assist mobile robot teleoperation within a shared autonomy system for remote task executions, this paper reports a novel approach to recognize the contextual task the human operator performs, by employing Sparse Online Gaussian Process to learn and classify human motion patterns executing various task types from demonstrations, due to its superior introspective capability over other state-of-art classification methods, such as Support Vector Machine (SVM), which is probably the most widely used approach on this topic to date. Our approach is evaluated on real data and shown to outperform current methods both in classification accuracy and uncertainty estimation regarding the predictive class labels, while maintaining sparsity to scale with large datasets.