- Adversarial Object Rearrangement in Constrained Environments with Heterogeneous Graph Neural Networks(arXiv)
Authors: Xibai Lou, Houjian Yu, Ross Worobel, Yang Yang, Changhyun Choi
Abstract: Adversarial object rearrangement within the precise world (e.g., beforehand unseen or outsized objects in kitchens and outlets) might acquire benefit from understanding job scenes, which inherently entail heterogeneous elements comparable to current objects, goal objects, and environmental constraints. The semantic relationships amongst these elements are distinct from each other and important for multi-skilled robots to hold out successfully in regularly eventualities. We propose a hierarchical robotic manipulation system that learns the underlying relationships and maximizes the collaborative vitality of its numerous experience (e.g., pick-place, push) for rearranging adversarial objects in constrained environments. The high-level coordinator employs a heterogeneous graph neural group (HetGNN), which causes in regards to the current objects, goal objects, and environmental constraints; the low-level 3D Convolutional Neural Group-based actors execute the movement primitives. Our technique is expert solely in simulation, and achieved a imply success price of 87.88% and a planning value of 12.82 in real-world experiments, surpassing all baseline methods. Supplementary supplies is obtainable at https://sites.google.com/umn.edu/versatile-rearrangement.
