Multi-Agent Collaboration for Space Mission Planning Using Agent-to-Agent Protocol

wix · Tech · 2026 · source Agentic: yes · promotional bias: med · source-backed · extracted with prompt 0.5.6 (schema v2)

Generated from extraction.json — do not edit by hand. Every claim cites verbatim source evidence; stated means the quote directly asserts the claim, implied means it supports it indirectly.

Problem

Domain: Experimental space mission planning

Wix's AI Core Architecture team wanted to test whether autonomous agents could collaborate on an open-ended optimization problem that no single agent could solve alone.

Constraints:

• Space mission planning must balance trajectory, launch windows, fuel, payload, crew, and timeline constraints.
• The demonstration intentionally avoided human intervention inside the agent collaboration loop.
• The system was experimental rather than a production mission-planning deployment.
• Agents needed to discover each other and negotiate dynamically rather than follow a fixed train track.

Architecture

The demonstration used three specialized agents: a Mission Coordinator for objectives and constraints, a Trajectory Agent for orbital mechanics, and a Resource Agent for fuel, payload, life support, and capacity constraints. They communicated through an Agent-to-Agent protocol that supported capability discovery, natural-language negotiation, selective engagement, and dynamic workflow generation.

Diagram confidence: reconstructed (reconstructed from prose).

flowchart LR
  a2a-protocol["A2A Protocol"] --> mission-coordinator["Mission Coordinator"]
  a2a-protocol --> trajectory-agent["Trajectory Agent"]
  a2a-protocol --> resource-agent["Resource Agent"]
  mission-coordinator --> trajectory-agent
  mission-coordinator --> resource-agent
  trajectory-agent --> resource-agent

Components

Data flow

The Mission Coordinator states mission objectives and constraints, discovers the trajectory and logistics agents through the A2A protocol, and asks them to evaluate feasibility. The Trajectory Agent computes transfer options and delta-v requirements while the Resource Agent checks fuel and payload constraints. When constraints conflict, the relevant agents negotiate directly until they find a feasible compromise.

Why this architecture

The problem was chosen because it forces conflicting specialist constraints rather than a single linear calculation. A2A let agents expose their capabilities, engage only when needed, and generate conversation flows dynamically, which demonstrated collaboration beyond predetermined workflows.

Models used: unspecified AI agents

Scale

• The Jupiter mission optimization reduced fuel requirements from 14.44 to 11.83 delta-v units.
  “reducing fuel requirements from 14.44 to 11.83 delta-v units—a solution not pre-programmed by the developers.”summary, stated
• Earth and Mars transfer windows align about every 26 months.
  “Planets align optimally for transfer only at specific intervals (Earth and Mars align every 26 months)”summary, stated
• The demonstration used three specialized agents.
  “The system consists of three specialized agents, each representing a distinct domain of expertise:”summary, stated

Alternatives considered (roads not taken)

Solve the planning problem with one agent

Rejected because: The experiment framed the problem as one that none of the specialist agents could solve independently.

Use predetermined workflows

Rejected because: The demonstration was meant to show agents adapting to unexpected situations through dynamic conversation.

Route every issue through the coordinator

Rejected because: Selective engagement lets conflicting specialists negotiate directly without polluting the whole context.

Patterns

multi-agent-orchestration (architecture)

The system splits mission planning across coordinator, trajectory, and logistics/resource specialists.

tool-use-orchestration (architecture)

Specialist agents expose capabilities and use domain functions such as trajectory optimization and feasibility analysis.

context-engineering (coordination)

Selective engagement limits negotiation to the agents with conflicting concerns, reducing unnecessary context pollution.

Tradeoffs

Use autonomous agent negotiation

Chose Dynamic A2A conversation among specialists over A fixed deterministic workflow.

Rationale: The demonstration sought solutions not explicitly coded by developers.

Let specialist agents negotiate directly

Chose Selective specialist engagement over Coordinator-mediated discussion for every conflict.

Rationale: Direct negotiation reduces context pollution and keeps the conversation focused.

Gotchas (what bit them)

A2A demonstrations can look more general than the evidence supports.

Root cause: The case is an experiment in a simulated mission-planning domain, not a production deployment.

Resolution: Treat it as evidence for collaboration patterns, not operational readiness in aerospace.

Open-ended agent collaboration can create context noise.

Root cause: If every agent participates in every negotiation, irrelevant concerns pollute context and consume tokens.

Resolution: The A2A pattern uses selective engagement so only conflicting specialists negotiate in detail.

Evaluation

The experiment evaluated whether agents could find feasible mission plans and negotiate around infeasible constraints; it is a qualitative demonstration rather than a production benchmark.

Metrics: 14.44 to 11.83 delta-v reduction · three specialized agents · 26-month Earth-Mars alignment window

Outcomes

• The agents found a gravity-assist solution for an initially infeasible Jupiter mission.
• The demo showed A2A discovery, negotiation, and dynamic workflow generation.
• The result is evidence of collaborative agent behavior, not validated production aerospace software.

The notable quantified result is reduction from 14.44 to 11.83 delta-v units in the Jupiter scenario.