Researcher proposes split edge-cloud compute architecture for planetary geology robot swarms

A developer explores how autonomous robot swarms could survey planetary geology while accounting for long communication delays and limited bandwidth. The author notes that surface missions like Perseverance cover only a small area, so distributed swarms of agents and orbital relays are proposed to map terrain more broadly. The work also considers linking compute to carbon-negative infrastructure, arguing that the energy balance between onboard inference and bulk data downlink matters. The author built a simulator suggesting the optimal edge-versus-cloud split is dynamic, depending on link quality, battery state, and ground-segment carbon intensity. The article presents a coordination architecture, edge inference stack, and carbon accounting layer, though some open problems remain unresolved.