spacemind-a-modular-and-self-evolving-embodied-vis

name: spacemind-a-modular-and-self-evolving-embodied-vis description: 'Research paper: SpaceMind: A Modular and Self-Evolving Embodied Vision-Language Agent Framework ' metadata: source: arXiv arxiv_id: 2604.14399 published: 2026-04-15 utility_score: 1.0 keywords: self-evolving, tool, tools, reasoning, robustness, autonomous, fine-tuning

SpaceMind: A Modular and Self-Evolving Embodied Vision-Language Agent Framework for Autonomous On-orbit Servicing

arXiv ID: 2604.14399
Published: 2026-04-15
Utility Score: 1.0
URL: http://arxiv.org/abs/2604.14399

Authors

Aodi Wu, Haodong Han, Xubo Luo

Abstract

Autonomous on-orbit servicing demands embodied agents that perceive through visual sensors, reason about 3D spatial situations, and execute multi-phase tasks over extended horizons. We present SpaceMind, a modular and self-evolving vision-language model (VLM) agent framework that decomposes knowledge, tools, and reasoning into three independently extensible dimensions: skill modules with dynamic routing, Model Context Protocol (MCP) tools with configurable profiles, and injectable reasoning-mode skills. An MCP-Redis interface layer enables the same codebase to operate across simulation and physical hardware without modification, and a Skill Self-Evolution mechanism distills operational experience into persistent skill files without model fine-tuning. We validate SpaceMind through 192 closed-loop runs across five satellites, three task types, and two environments, a UE5 simulation and a physical laboratory, deliberately including degraded conditions to stress-test robustness. Under nominal conditions all modes achieve 90--100% navigation success; under degradation, the Prospective mode uniquely succeeds in search-and-approach tasks where other modes fail. A self-evolution study shows that the agent recovers from failure in four of six groups from a single failed episode, including complete failure to 100% success and inspection scores improving from 12 to 59 out of 100. Real-world validation confirms zero-code-modification transfer to a physical robot with 100% rendezvous success. Code: https://github.com/wuaodi/SpaceMind

Matched Keywords

self-evolving, tool, tools, reasoning, robustness, autonomous, fine-tuning

Relevance to AI Agents

This paper is highly relevant to AI agent systems research with focus on:

self-evolving, tool, tools, reasoning, robustness

Quick Reference

# View paper
open http://arxiv.org/abs/2604.14399

# Download PDF
open http://arxiv.org/pdf/2604.14399.pdf

Auto-generated from arXiv on 2026-04-17

Activation Keywords

"spacemind-a-modular-and-self-evolving-embodied-vis"
"spacemind a modular and self evolving embodied vis"
"use spacemind a modular and self evolving embodied vis"
"spacemind a modular and self evolving embodied vis help"
"spacemind a modular and self evolving embodied vis tool"

Tools Used

Read - Read existing files and documentation
Write - Create new files and documentation
Bash - Execute commands when needed

Instructions for Agents

Identify user's intent and specific requirements
Gather necessary context from files or user input
Execute appropriate actions using available tools
Provide clear results and suggest next steps

Examples

Basic Spacemind A Modular And Self Evolving Embodied Vis usage

User: "Help me with spacemind a modular and self evolving embodied vis"
→ Understand requirements → Execute actions → Provide results

Advanced usage

User: "I need detailed spacemind a modular and self evolving embodied vis assistance"
→ Clarify scope → Provide comprehensive solution → Follow up