Active Visuo-Tactile Interactive Robotic Perception for Accurate Object Pose Estimation in Dense Clutter

Deep Active Cross-Modal Visuo-Tactile Transfer Learning for Robotic Object Recognition

“…We perform Markov Monte-Carlo sampling of N viewpoints on the hemisphere space located above the centroid o centroid of the bounding box of the object of interest which is known a priori. The 3D position p view is randomly sampled as a point on the hemisphere and the orientation of the view as axis of rotation ⃗e and angle θ is computed with [34]:…”

“…Deep Active Visual Object Learning: If the scene is cluttered, we use our prior work in Murali et al [34] to autonomously declutter the workspace. After the scene is decluttered, the robots initiate visual data collection.…”

“…Furthermore, contrasting to visual perception, tactilebased recognition requires interaction with the objects as data is collected upon contact with the objects [8]. To reduce redundant data collection, temporal costs and human intervention, several approaches have been proposed for performing active data acquisition through information-gain based action selection [10]- [12], [32]- [34]. Leveraging active perception and learning techniques can aid in reducing data collection costs and improve time efficiency for vision-to-tactile crossmodal domain adaptation.…”

An Empirical Evaluation of Various Information Gain Criteria for Active Tactile Action Selection for Pose Estimation

“…Similarly, Shannon entropy has been used for selecting actions that provides the maximum discriminatory information for object classification [10], [19]. In our previous works [7], [8] KL-divergence has been used for tactile action selection for object pose estimation.…”

Towards Robust 3D Object Recognition with Dense-to-Sparse Deep Domain Adaptation

“…However, LiDAR-based point clouds are very sparse with increasing distance, ranging between 10-100 points [1], [2]. Such sparse point clouds are also produced with commercial tactile sensors [3]- [5] (Figure 1). Most state-of-art methodologies rely on relatively dense point clouds and the performance drops drastically with sparse point clouds [6].…”