Logo image
Back
Adaptive AI-Driven, Edge-Computing Robotic Platforms for Advanced Human-Robot Collaboration in Extraterrestrial Habitat Construction
Conference paper

Adaptive AI-Driven, Edge-Computing Robotic Platforms for Advanced Human-Robot Collaboration in Extraterrestrial Habitat Construction

Nijanthan Vasudevan, Oudayl Massat, Arjuna Karthikeyan Senthilvel Kavitha, Padmapriya Sampathkumar and Pavithiran Vasudevan
IAA Symposium on Human Exploration of the Solar System, pp 277-288
2025
DOI: https://doi.org/10.52202/083078-0028

Abstract

The establishment of sustainable human habitats on the Moon and Mars necessitates a paradigm shift in construction methodologies, moving from direct human labor towards advanced human-robot collaboration. The extreme environmental challenges-including vacuum, significant thermal fluctuations, abrasive dust, and communication latencies-render traditional construction and teleoperation models inefficient and hazardous. This paper introduces the Adaptive AI-Driven Robotic Assistant (AIDRAS), an integrated robotic platform designed to function as a capable collaborator for astronauts in extraterrestrial surface construction missions. AIDRAS leverages a synergistic combination of edge computing for low-latency decision-making, deep reinforcement learning (DRL) for mastering complex manipulation tasks, and a novel federated learning (FL) framework for decentralized, collaborative skill acquisition across a multi-robot team. This approach mitigates the operational constraints imposed by communication delays and bandwidth limitations. The system architecture is founded on a modular hardware platform with interchangeable end-effectors, enabling diverse tasks from structural assembly to non-destructive inspection. Human-Robot Interaction (HRI) is facilitated through an intuitive interface featuring augmented reality (AR) overlays and natural language processing, allowing astronauts to supervise tasks at a high level of abstraction, thereby reducing cognitive load and enhancing mission safety. We present a comprehensive mathematical framework for the core algorithms, including a graph-based formulation for robust state estimation and a theoretical treatment of the federated learning process. The platform's capabilities are validated through a pipeline of high-fidelity, physics-based simulations in ROS/Gazebo and hardware-in-the-loop experiments, demonstrating robust performance in representative construction scenarios. This research culminates in an AI-driven robotic solution that adapts to dynamic mission requirements, reduces operational costs, and accelerates the critical path towards establishing a permanent human presence beyond Earth. © 2025 by the International Astronautical Federation (IAF).

Metrics

3 Record Views

Details

Logo image