Federated Multi-Agent Mapping for Planetary Exploration

2024-04-02

arXiv.org

AI Summary

To address multi-robot collaborative mapping under severe bandwidth constraints in deep-space exploration, this paper proposes a federated multi-agent implicit neural mapping (INM) framework: raw sensor data are never transmitted; instead, only model parameters are exchanged among agents for on-orbit joint training of a global implicit neural representation (INR) map. We innovatively incorporate Earth-derived geometric priors via meta-learning initialization, reducing convergence iterations by 80% and achieving a 93.8% map compression ratio. A traversability-aware transfer training strategy further enhances cross-terrain generalization. Evaluated on Mars terrain and glacial datasets, our method achieves an F1-score of 0.95 for path planning and significantly outperforms existing baselines in map reconstruction accuracy. This work pioneers the tight integration of federated learning, implicit neural representations, and meta-initialized optimization—establishing a novel, resource-efficient, compact, and collaborative online mapping paradigm tailored for bandwidth- and compute-constrained space missions.

Technology Category

• Intelligent Robots: Localization, Mapping, and Navigation
• Multiagent Systems: Multiagent Learning
• Machine Learning: Distributed Machine Learning & Federated Learning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Multi-agent robotic exploration stands to play an important role in space exploration as the next generation of robotic systems ventures to far-flung environments. A key challenge in this new paradigm will be to effectively share and utilize the vast amount of data generated onboard while operating in bandwidth-constrained regimes typical of space missions. Federated learning (FL) is a promising tool for bridging this gap. Drawing inspiration from the upcoming CADRE Lunar rover mission, we propose a federated multi-agent mapping approach that jointly trains a global map model across agents without transmitting raw data. Our method leverages implicit neural mapping to generate parsimonious, adaptable representations, reducing data transmission by up to 93.8% compared to raw maps. Furthermore, we enhance this approach with meta-initialization on Earth-based traversability datasets to significantly accelerate map convergence; reducing iterations required to reach target performance by 80% compared to random initialization. We demonstrate the efficacy of our approach on Martian terrains and glacier datasets, achieving downstream path planning F1 scores as high as 0.95 while outperforming on map reconstruction losses.

Problem

• Efficient data sharing in bandwidth-constrained space missions.
• Federated learning for multi-agent mapping without raw data transmission.
• Accelerating map convergence using Earth-based datasets for planetary exploration.

Innovation

• Federated learning for multi-agent map sharing
• Implicit neural mapping reduces data transmission
• Meta-initialization accelerates map convergence