Authors: Mario A.V. Saucedo, Akash Patel, Akshit Saradagi, Christoforos Kanellakis, George Nikolakopoulos

Published on: February 06, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2402.03840

Summary

• What is new: Introduction of Belief Scene Graphs (BSG) for efficient high-level task planning in robotics with partial information.
• Why this is important: The challenge of task planning in robotics with incomplete 3D information.
• What the research proposes: A graph-based learning method called Computation of Expectation based on Correlation Information (CECI) for approximating real belief on 3D scene graphs.
• Results: The CECI model, trained on a novel dataset of 3D scene graphs, effectively enables robots to plan tasks by integrating expectations into abstract representations.

Technical Details

Technological frameworks used: Graph Convolutional Neural Network (GCN)

Models used: CECI model for learning from 3D scene graphs.

Data used: A new dataset of 3D scene graphs generated from semantically annotated real-life 3D spaces.

Potential Impact

Robotics, autonomous vehicles, augmented reality companies, and markets could significantly benefit or face disruption from these insights.

Want to implement this idea in a business?

We have generated a startup concept here: GraphSense.