Autonomous Reconstruction of Unknown Indoor Scenes Guided by Time-varying Tensor Fields

ACM Transactions on Graphics (Proceedings of SIGGRAPH Asia 2017) 

Kai Xu1,2     Lintao Zheng2    Zihao Yan1   Guohang Yan  Eugene Zhang3   Matthias Niessner4    Oliver Deussen5   Daniel Cohen-Or1,6    Hui Huang1*

1Shenzhen University     2National University of Defense Technology     3Oregon State University    4Stanford University    5University of Konstanz    6Tel-Aviv University

Figure 1: We present an autonomous system for active object identification in an indoor scene (a), with consecutive depth acquisitions, for online scene modeling. The scene is first roughly scanned, and segmented to generate 3D object proposals. Targeting an object proposal (b), the robot performs multi-view object identification, based on a 3D shape database, driven by a 3D Attention Model. The retrieved 3D models are inserted into the scanned scene (c), replacing the corresponding object scans, thus incrementally constructing a 3D scene model (d).

Autonomous reconstruction of unknown scenes by a mobile robot inherently poses the question of balancing between exploration efficacy and reconstruction quality. We present a navigation-by-reconstruction approach to address this question, where moving paths of the robot are planned to account for both global efficiency for fast exploration and local smoothness to obtain high-quality scans. An RGB-D camera, attached to the robot arm, is dictated by the desired reconstruction quality as well as the movement of the robot itself. Our key idea is to harness a time-varying tensor field to guide robot movement, and then solve for 3D camera control under the constraint of the 2D robot moving path. The tensor field is updated in real time, conforming to the progressively reconstructed scene. We show that tensor fields are well suited for guiding autonomous scanning for two reasons: first, they contain sparse and controllable singularities that allow generating a locally smooth robot path, and second, their topological structure can be used for globally efficient path routing within a partially reconstructed scene. We have conducted numerous tests with a mobile robot, and demonstrate that our method leads to a fluent exploration and high-quality reconstruction of unknown indoor scenes.

Figure 2: An overview of our method and system. Our system runs an online scene reconstruction and employs an occupancy map for storing spatial occupancy information (a). The progressively reconstructed 3D scene geometry is projected onto the floor plane (b-left), to compute a geometry-aware time-varying tensor fields. Robot movement is locally directed by path advection over the fields (b-middle), and globally guided with path finding, based on the field topology (b-right). A smooth camera trajectory is computed along the path (c).

Figure 3: The topological skeleton of tensor field can be computed for a partially scanned scene (a) and used for guiding the robot scanning. When the robot (white dot) arrives at a trisector, a minimum cost spanning tree is generated from the topological graph, to enable branch selection (b). When the reconstruction is complete, the field topology (c) conforms approximately to the medial axis of the full scene boundary (d).

Figure 4: Four real scenes scanned and reconstructed by our autonomous system. For each scene, we show the final field topology (left) and the reconstruction result (right). The scene in (c) is not closed, due to inaccessible narrow doors; the scanning was terminated by human..


Download and Reference

We have released our source code built on top of the ROS, and contributed a software package to ROS.

ROS package of tensor_field_nav

[To reference our algorithm, code or data in a publication, please include the bibtex below and a link to this website.]

@article {Fetch1,
    title = {Autonomous Reconstruction of Unknown Indoor Scenes Guided by Time-varying Tensor Fields},
    author = {Kai Xu and Lintao Zheng and Zihao Yan and Guohang Yan and Eugene Zhang and Matthias Niessner and Oliver Deussen and Daniel Cohen-Or and Hui Huang},
    journal = {ACM Transactions on Graphics (Proc. of SIGGRAPH Asia 2017)},
    volume = {36},
    number = {6},
    pages = {202:1-15},
    year = {2017}
Copyright © 2016-2018 Visual Computing Research Center