Edge-to-Cloud Swarm Coordination for planetary geology survey missions in carbon-negative infrastructure

说明与用途

一种层次化的 edge‑to‑cloud 策略，用于行星级机器人群体协调，在边缘推理、轨道融合和云端规划之间平衡计算，并考虑碳强度。

问题

How to coordinate a planetary swarm of surface agents, orbital relays, and ground/cloud resources to achieve global geological mapping while managing limited edge compute, intermittent connectivity, and variable carbon intensity of the ground segment.

结果

The dynamic edge-to-cloud policy improves science value by aligning decisions across tiers, reduces energy consumption on surface agents, and enables carbon-aware scheduling that adapts to link and ground conditions.

工作流

• 1. Distill a cloud-trained planner into a ~2M parameter edge policy with value head using value-aware distillation.
• 2. Deploy the distilled policy on surface agents with a few watts compute budget.
• 3. Equip orbital relays with higher power to run mid-sized models for task allocation and data fusion.
• 4. Ground/cloud segment hosts global geology models, simulation-based planning, and policy training.
• 5. Implement a carbon-aware scheduler that balances edge inference and bulk downlink based on real-time link metrics and ground carbon intensity.
• 6. Share a common latent science-value representation across tiers to align objectives.

证据

I distilled a cloud-trained planner into a ~2M parameter network that runs in under 50ms on an edge accelerator.

来源

The tiers should share a common latent representation of "science value" so that decisions at any tier can be evaluated against the same objective.

来源