Welcome to the ICRA 2024 Workshop on Neural Fields in Robotics!

This workshop aims to explore the role and potential of Neural Fields i.e. in various robotics domains, including 6D object pose estimation, SLAM, manipulation with reinforcement learning (RL), object reconstruction, neural implicit data generation, few-view scene reconstruction, camera calibration, physics modeling, and planning/navigation. By leveraging recent advancements in computer vision, such as neural radiance fields (NeRFs) and deep Signed Distance Functions (DeepSDFs), this workshop aims to foster discussions and collaborations in the robotics community.

Through invited and selected spotlight talks as well as a room for discussions, the workshop will address several key topics:

• The advantages of neural fields beyond reconstruction for the robotics context
• Trade-offs between an implicit and explicit representation
• Online training of neural fields and the speed/efficiency trade-offs
• Balancing fully-implicit representations which are less interpretable with hybrid neural field based representations which offer more interpretability

By facilitating knowledge exchange and exploration, we hope the workshop sparks interest in neural fields for robotics, shaping the future of the field and driving innovation in perception, planning, and control systems.

📚 Topics

The RoboNerF workshop will accept papers in relation to the application of neural fields and implicit representation to robotics areas. Topics of this workshop include but are not limited to:

• Neural Fields for Real-time Robotics Applications
• Gaussian Splatting (Explicit) and other hybrid representations i.e. Triplanes for Robotics Applications
• NeRFs and Implicit Representations for SLAM
• Implicit Grasping for Manipulation
• 6D Pose Estimation using Implicit Fields
• Mobile Robot Navigation using Implicit Functions
• Efficient Hybrid Representations for 3D Scene Understanding
• Articulated Object Reconstruction and Manipulation
• Neural Implicit Data Generation and Capture of 3D Scenes
• Tactile and Deformable Object Manipulation Using Learnt Implicit Models
• Physical and Dynamical Properties Estimation Based on Neural Field Models
• ...

📨 Submission

Submissions should follow ICRA format (i.e. IEEE conference template) and not exceed 4 pages including figures and tables but excluding references and potential supplementary material. The submitted abstracts will be peer-reviewed in ICRA format (single-blind). Accepted abstracts will be made non-archival public through this website, and a portion of the accepted papers will be orally presented at the workshop. Any type of contributions, either already published (or accepted) or works in progress, are warmly welcome.

Paper submissions will be handled with OpenReview through the following link:

RoboNerF @ OpenReview