In-Situ Resource Utilization (ISRU)

In-situ resource utilization  is the practice of building with what is already at the destination — regolith, water ice, volatiles — rather than hauling it from Earth. It is not an optimization. Large bases, sustained crewed presence, and interplanetary supply chains are simply infeasible while the mass ratio is dominated by Earth-launched feedstock, because the rocket equation is unsympathetic.

The mining, refining, and construction are hard. So is the coordination, and it gets less attention.

Why swarms

ISRU concepts have been dominated for two decades by monolithic bespoke machines: Curiosity-class excavators, large printers, specialized refineries. Per unit of mass they are efficient. They are also poorly matched to the operating regime, for three reasons.

Throughput scales with parallelism, not with individual size. Thousands of small agents working regolith simultaneously outperform one large machine of comparable total mass, because excavation is embarrassingly parallel and a single machine’s duty cycle is bounded by its own power and thermal limits.

Resilience requires multiplicity. A single excavator that fails halts the base, and there is no repair crew. A swarm losing 10% of its agents still delivers roughly 90% of its throughput, and it degrades continuously rather than catastrophically. This is distributed robustness in its most literal form.

Adaptability requires decentralization. Regolith composition, slope, and consolidation vary over metres. A global schedule computed from an orbital survey is wrong at the scale at which the work happens; local policies acting on local sensing are not.

A worked example at the scale of a lunar landing pad is in Energy-Aware Hierarchical Markets, measured over 20 seeds per cell. It carries a precondition worth reading before adopting the mechanism: the auction beats a first-come baseline only where hauling capacity is scarce, and when haulers are over-provisioned it is measurably worse than doing the simplest thing.

Macro–micro coordination

Base-scale ISRU needs two tiers of agent that must stay in step despite operating at different scales, rates, and energy budgets.

Micro agents are millimetre- to centimetre-scale: excavation, compaction, sintering precursors. They are severely energy-constrained, have short communication range, fail individually and often, and act on local rules alone.

Macro agents are rovers, haulers, printers, and charging stations at kilogram to tonne scale. They travel further, carry real energy budgets, and can execute structured schedules.

Coupling them is the interesting problem, because their production and consumption rates do not match and cannot be made to. The mechanism we study uses a market with energy-aware bids and buffered depots: buffers decouple the macro and micro rate mismatch, so neither tier stalls waiting for the other; energy-aware bids prevent a depot from draining its haulers into a brownout from which nothing can recover; and relay rotation stops any single charging station from becoming a point of failure the whole base depends on.

Open research questions

Optimal fleet composition. Given a target production rate and an energy budget, what is the right ratio of micro to macro agents? Currently set empirically. It belongs in a benchmark suite rather than in each paper’s appendix.

Depot geometry. How many depots, at what separation, for a base of a given size? Depot count interacts nonlinearly with hauler count and total fleet energy, and the interaction is where the interesting failures live.

Honest energy accounting. Published mobility-energy figures understate thermodynamic cost by one to two orders of magnitude, because sintering lunar regolith costs roughly 1–5 MJ/kg and mobility does not. Any ISRU energy model that omits process heat is measuring the wrong thing. See energy efficiency and resource constraints.

Wear propagation. Agents are modelled as failing at a random rate, independently. Real machines fail from accumulated mechanical wear, correlated with the work they have done and with the abrasiveness of the regolith they have done it in. A wear-correlated failure model would change fleet-composition tradeoffs materially, and probably not in the direction the current models suggest.

Coordination under distance. Agents spread across a base site, with duty cycles governed by power availability and line-of-sight interrupted by terrain, do not observe each other promptly. Our measurement is that coordination quality collapses through a boundary once delay grows relative to the task timescale — so whether a lunar ISRU fleet coordinates at all depends on a ratio nobody in this literature currently reports.

Adjacent work

Classical ISRU surveys by Sanders and Larson, and the IEEE Aerospace proceedings, cover the process engineering. Non-swarm robotic constructors include NASA’s RASSOR excavator and MIT’s TERRA project. Market-based multi-robot coordination traces to Dias, Stentz, and Parker, and the multi-robot task allocation taxonomy that followed.

Arboria’s interest is the intersection: heterogeneous macro–micro fleets, energy-aware market scheduling, and what happens to all of it when the agents cannot see each other in real time.

Related

Planetary-scale engineering sets the broader context. Autonomous manufacturing systems covers the terrestrial analogue. Orbital compute constellations is the same coordination problem where the delay is imposed by light rather than by terrain.