Learning spatial relationships between objects

Abstract

Although a manipulator must interact with objects in terms of their full complexity, it is the qualitative structure of the objects in an environment and the relationships between them which define the composition of that environment, and allow for the construction of efficient plans to enable the completion of various elaborate tasks. In this paper we present an algorithm which redescribes a scene in terms of a layered representation, from labeled point clouds of the objects in the scene. The representation includes a qualitative description of the structure of the objects, as well as the symbolic relationships between them. This is achieved by constructing contact point networks of the objects, which are topological representations of how each object is used in that particular scene, and are based on the regions of contact between objects. We demonstrate the performance of the algorithm, by presenting results from the algorithm tested on a database of stereo images. This shows a high percentage of correctly classified relationships, as well as the discovery of interesting topological features. This output provides a layered representation of a scene, giving symbolic meaning to the inter-object relationships useful for subsequent commonsense reasoning and decision making.

Learning Kinematic Structure Correspondences Using Multi-Order Similarities

“…This corresponds to the minimum and maximum range of all possible combinatorial motions of 3. There have been many works trying to find joint positions as part of kinematic structure estimation [13], [14], [15], [71]. Yan and Pollefeys [13] estimate joint positions as the intersection of two motion subspaces, but their method is vulnerable to noise.…”

“…The estimate in [14] is based on probabilistic inference which can find the number of joints and their connections adaptively, but the method is sensitive to the prior and has difficulty in recovering from a poor initialisation. In [71] a SVM based method which is conceptually similar to our proposal is presented, but ours is simpler than theirs. We tested these methods, and have found that their joint estimates are similar or worse compared to ours.…”

Support relation analysis and decision making for safe robotic manipulation tasks

“…Rosman and Ramamoorthy [25] present a method for learning spatial relationships between objects using the segmented point clouds. In their work the lack of information about the complete set of objects and physical interactions between objects are not considered.…”

Predicting object interactions from contact distributions

“…Rosman and Ramamoorthy [21] proposed the use of a contact network to learn the spatial relations between objects. Contact points were detected using a support vector machine to separate the point clouds of the objects.…”