Efficient Distributed Semantic-Relational Collaborative SLAM for Structured Environments
Multi S-Graphs is a decentralized CSLAM system that utilizes high-level semantic-relational information embedded in a four-layered hierarchical and optimizable situational graph to enable cooperative map generation and localization in structured environments while minimizing the information exchanged between robots.
Efficient Large-Scale Incremental Dense Mapping using Robot-centric Implicit Neural Representation
The proposed Robot-centric Implicit Mapping (RIM) technique enables efficient, scalable, and high-quality large-scale incremental dense mapping using range sensors by leveraging a robot-centric local map, a decoupled global map, bundle supervision, and outlier removal.
Lightweight Open-Set Semantic Simultaneous Localization and Mapping (LOSS-SLAM)
A computationally efficient method for open-set semantic localization and mapping that utilizes self-supervised vision transformer features (DINO) to augment geometric correspondence matching at the object level.
Federated Learning for Efficient and Adaptive Multi-Agent Mapping in Planetary Exploration
A federated learning approach for collaborative multi-agent mapping that leverages meta-initialized 2D neural radiance fields to enable rapid adaptation and efficient data sharing across diverse planetary environments.
© 2024 by Linnk AI