[en] Graphs [1] presented a novel factor graph compounded by robot keyframes associated with a scene graph that includes higher-level semantic-relational concepts. However, it only includes simple and rigid room structures and lacks other useful entities such as walls (as a wall surface to wall surface relationship). To overcome this, we propose to generate new factorized subgraphs using the relational and generalisation power of Graph Neural Networks (GNN)
Disciplines :
Computer science
Author, co-author :
MILLÁN ROMERA, José Andrés ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Automation
BAVLE, Hriday ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Automation ; I. MOTIVATION To enhance Simultaneous Localization and Mapping
VOOS, Holger ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Automation ; I. MOTIVATION To enhance Simultaneous Localization and Mapping
SANCHEZ LOPEZ, Jose Luis ; University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > Automation
External co-authors :
no
Language :
English
Title :
Late breaking results on Graph Reasoning on Situational Graphs for higher-concepts generation
Publication date :
October 2023
Event name :
2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS'23)
