Spatial Relation Graph and Graph Convolutional Network
for Object Goal Navigation
IEEE CASE 2022
- D. A. Sasi Kiran* IIIT Hyderabad, India
- Kritika Anand* TCS Research, India
- Chaitanya Kharyal* IIIT Hyderabad, India
- Gulshan Kumar IIIT Hyderabad, India
- Nandiraju Gireesh IIIT Hyderabad, India
- Snehasis Banerjee TCS Research, India
- Ruddra dev Roychoudhury TCS Research, India
- Mohan Sridharan University of Birmingham, UK
- Brojeshwar Bhowmick TCS Research, India
- Madhava Krishna IIIT Hyderabad, India
Abstract
This paper describes a framework for the object-goal navigation (ObjectNav) task, which requires a robot to find and move to an instance of a target object class from a random starting position. The framework uses a history of robot trajectories to learn a Spatial Relational Graph (SRG) and Graph Convolutional Network (GCN)-based embeddings for the likelihood of proximity of different semantically-labeled regions and the occurrence of different object classes in these regions. To locate a target object instance during evaluation, the robot uses Bayesian inference and the SRG to estimate the visible regions, and uses the learned GCN embeddings to rank visible regions and select the region to explore next. This approach is tested using the Matterport3D (MP3D) benchmark dataset of indoor scenes in AI Habitat, a visually realistic simulation environment, to report substantial performance improvement in comparison with state of the art baselines.
Video
Citation
Acknowledgements
The website template was borrowed from Michaël Gharbi.