Agent Mars: Multi-Agent Simulation for Multi-Planetary Life Exploration and Settlement

TL;DR

Agent Mars provides a benchmarkable, auditable foundation for Space AI, an open, end-to-end multi-agent simulation framework for Mars base operations and proposes the Agent Mars Performance Index (AMPI), an interpretable composite score with diagnostic sub-metrics.

Abstract

Artificial Intelligence (AI) has transformed robotics, healthcare, industry, and scientific discovery, yet a major frontier may lie beyond Earth. Space exploration and settlement offer vast environments and resources, but impose constraints unmatched on Earth: delayed/intermittent communications, extreme resource scarcity, heterogeneous expertise, and strict safety, accountability, and command authority. The key challenge is auditable coordination among specialised humans, robots, and digital services in a safety-critical system-of-systems. We introduce Agent Mars, an open, end-to-end multi-agent simulation framework for Mars base operations. Agent Mars formalises a realistic organisation with a 93-agent roster across seven layers of command and execution (human roles and physical assets), enabling base-scale studies beyond toy settings. It implements hierarchical and cross-layer coordination that preserves chain-of-command while allowing vetted cross-layer exchanges with audit trails; supports dynamic role handover with automatic failover under outages; and enables phase-dependent leadership for routine operations, emergencies, and science campaigns. Agent Mars further models mission-critical mechanisms-scenario-aware short/long-horizon memory, configurable propose-vote consensus, and translator-mediated heterogeneous protocols-to capture how teams align under stress. To quantify behaviour, we propose the Agent Mars Performance Index (AMPI), an interpretable composite score with diagnostic sub-metrics. Across 13 reproducible Mars-relevant operational scripts, Agent Mars reveals coordination trade-offs and identifies regimes where curated cross-layer collaboration and functional leadership reduce overhead without sacrificing reliability. Agent Mars provides a benchmarkable, auditable foundation for Space AI.

Figures (7)

• Figure 1: Conceptual Mars base operations (left) and the Agent Mars roster of 93 agents organised into 7 layers of command and execution (right).
• Figure 2: Roster on Mars. 93 agents organized into 7 layers with 18 functional groups.
• Figure 3: 7-layer hierarchy.
• Figure 6: HCLC routing with curated shortcuts and audited hub-forwarding. (a) Solid edges ($\mathcal{E}_H$) implement the strict hierarchy. Dashed arcs are representative whitelist shortcuts $\mathcal{E}_X(\mathcal{W})$ consistent with the default code configuration (e.g., GEO$\rightarrow$COM, COM$\rightarrow$AI, and LSS/PWR/ISRU/AGRI/MNT$\rightarrow$AI). (b) When $(u,v)\notin\mathcal{W}$, traffic is forwarded via the mission hub (OPS) with audit logging.
• Figure 7: Scenario-aware memory timeline. Top: short-term window only ($M^{\text{short}}$ with horizon $k$) forgets early context ($Q1$–$Q2$), so later queries re‑ask earlier items (e.g., $Q2 \rightarrow Q5$). Bottom: adding distilled long‑term recall $f(M^{\text{long}})$ (and optional shared memory) enables state carry‑over, reducing re‑asks and rework; see Table \ref{['tab:memory']}.