The Missing Layer in Digital Twins for the Lunar Construction: Ground Behavior, Construction Reality, and Engineering Decisions

The Industry Assumption

Digital twins are being positioned as the backbone of future infrastructure. The expectation is straightforward: if models, data, and sensors are integrated into a single environment, we gain a continuously updated representation of reality, and with that, better design decisions, more efficient construction, and more reliable operations.

This assumption is not abstract. It is already shaping how infrastructure is being delivered. Across sectors, there is a consistent push to merge BIM and GIS into unified environments, stream real-time data from sensors and machines, and maintain a living model that reflects the state of the system as it evolves. The conclusion that follows is almost automatic: if the model is sufficiently detailed and continuously updated, the system becomes predictable.

That logic holds in many domains. It works for mechanical systems, electrical networks, and controlled processes where behavior is governed by well-defined parameters and where variability can be bounded with confidence. In those cases, the digital twin becomes a reliable extension of the physical system.

In infrastructure, the governing system is the ground. And the ground is not a boundary condition that can be defined once and carried through the design. It is not a static input that becomes more accurate with additional data. It is a material system with its own structure, history, and variability, and it responds to construction in ways that are not only uncertain but also continuously evolving.

As soon as construction begins, the system changes. Excavation alters stress fields. Traffic modifies density and fabric. Repeated loading accumulates deformation. Disturbance reshapes stiffness and strength. What is being modeled is not a fixed system that is gradually revealed with better data. It is a system that is being actively modified by the very processes we are trying to predict.

Digital twins, as currently conceived, rely on the idea that reality can be captured, updated, and converged through data integration. That premise depends on the underlying system being stable enough for that convergence to occur. When the governing system is changing as a function of construction itself, the problem is no longer one of resolution or data density.

The gap is subtle, but it is fundamental. The assumption is that more data and better models will close the distance between prediction and reality. In practice, the system is evolving faster than that distance can be reduced. The issue is not that we lack data. It is that the behavior we need to predict is being continuously redefined.

That distinction matters because it shifts the problem away from integration and toward engineering judgment. And that is where the rest of this discussion begins.

What Digital Twins Actually Do Today?

In practice, digital twins in infrastructure are not a single system. They are an integration of several mature tools that have been developed independently over decades and are now being connected under a common environment.

At the core, we have BIM and GIS providing the geometric and spatial backbone. BIM defines the asset in terms of components, interfaces, and specifications. GIS situates that asset within a broader spatial and environmental context. Around this, numerical models simulate behavior, structural response, groundwater flow, deformation, thermal effects, depending on the problem at hand. Instrumentation systems then feed observations back into the process, tracking settlement, pore pressure, displacement, vibration, and other measurable responses.

More recently, this ecosystem has been expanded with higher-frequency data sources. LiDAR and photogrammetry capture evolving geometry during construction. Machine telemetry provides operational parameters, loads, movements, and production rates. In some sectors, sensor networks stream near-real-time data into centralized platforms. The intent is to create a continuous loop where design, construction, and operation are informed by a shared and updated representation of the system.

This is a significant step forward compared to traditional workflows. Coordination has improved. Visibility has improved. The ability to detect deviations from expected performance has improved. In many cases, issues that would previously go unnoticed until late stages are now identified earlier.

What exists today is still a layered environment rather than a truly unified one. The design model, the construction model, and the monitoring data are connected, but they are not fully coupled. Updates often flow in one direction, data informing interpretation, rather than forming a closed, predictive loop where the model evolves with the system in a consistent and validated way.

More importantly, the integration is strongest where behavior is already well understood. Structural systems, mechanical components, and controlled processes benefit directly from this approach because their response can be defined within relatively stable boundaries. In those cases, the digital twin becomes a reliable extension of the physical system.

Geotechnical models are typically developed at the design stage, based on site investigation data, correlations, and assumptions about variability. During construction, instrumentation provides observations that are compared against predicted behavior. When deviations occur, the model may be updated, but this process is not continuous or systematically integrated into a predictive framework that evolves with each construction step.

In effect, the digital twin captures the geometry of what is being built and the measured response of the system, but it does not fully capture how the ground itself is changing as construction progresses.

This distinction is often masked by the level of sophistication in the tools. The models are advanced. The data streams are rich. The visualization is convincing. But the underlying coupling between ground behavior and system representation remains limited.

As a result, what is often referred to as a digital twin in infrastructure today is better described as a coordinated digital environment. It provides a clearer picture of the system, but it does not yet function as a fully predictive, continuously evolving representation of ground-governed behavior.

Where is the Industry Pushing the Limits?

If the goal is to understand how far current practice can be extended toward lunar and Martian construction, it is necessary to look at the sectors that already operate closest to a continuously changing ground environment.

Mining is one of them. Open-pit and underground operations have moved toward highly instrumented, data-rich workflows where geometry, material movement, and equipment performance are tracked at high frequency. Fleet management systems, LiDAR scanning, slope radar, and short-interval control provide a near-continuous picture of how the excavation evolves. There is a strong operational feedback loop, particularly in production-driven environments, where decisions are adjusted based on observed response.

Oil and gas go further in a different direction. Drilling operations rely on real-time subsurface data through Measurement While Drilling (MWD) and Logging While Drilling (LWD). Geosteering integrates these measurements into evolving subsurface interpretations, allowing operators to adjust trajectories as conditions change. The model is not static; it is updated during execution, and decisions are made within that loop.

Measuring While Driving (MWD) in Tunnel Boring Machine (TBM) Excavations in Earth-based projects.

Mechanized underground excavation, particularly with Tunnel Boring Machines and raise boring systems, introduces another layer. These machines generate continuous streams of data, torque, thrust, penetration rate, and cutterhead response that are directly influenced by ground conditions. MWD data in raise boring, and operational parameters in TBMs, are routinely used to infer changes in rock mass behavior along the alignment. In effect, the excavation process itself becomes a sensing mechanism.

These three domains share a common characteristic: they operate in environments where the ground is being continuously modified, and where decisions must adapt to that reality.

But even here, the limitations are clear.

The data is rich, but the interpretation remains indirect. MWD, machine telemetry, and drilling responses provide proxies of ground conditions, not direct state definitions. The models evolve, but they do not update deterministically with each excavation or drilling step. Disturbance effects, loosening, densification, and fabric alteration are not rigorously captured as part of a unified system. Uncertainty is acknowledged, but it is not consistently propagated into engineering decisions in a structured way.

In other words, these industries are not operating true digital twins of the ground. They are operating advanced observational systems with strong feedback loops. That distinction becomes critical when considering off-planet construction. On Earth, these limitations are often manageable. There is tolerance for adjustment. Additional investigation can be performed. Construction sequences can be modified. Equipment can be replaced or reinforced. The system can absorb uncertainty through operational flexibility.

On the Moon or Mars, that margin is significantly reduced.

Operations will be constrained by logistics, energy, autonomy, and limited intervention capability. The ability to adapt in the field will exist, but it will be far more restricted than on Earth. At the same time, the ground conditions introduce additional complexities, low confinement, vacuum, thermal cycling, and regolith behavior, that are not directly analogous to terrestrial environments. This creates a situation where the current frontier of Earth-based practice is not sufficient, but it is highly instructive.

Mining, oil and gas, and mechanized excavation show that continuous interaction with the ground can be measured, interpreted, and partially integrated into decision-making. They also show that even under those conditions, the ground is not yet represented as a fully evolving, predictive system within a digital twin framework.

For lunar and Martian construction, that is the level that needs to be exceeded, not replicated.

The question is no longer how to integrate more data into the model. It is how to represent a ground system that evolves with construction, under conditions where uncertainty cannot be managed through trial and adjustment alone.

That is where the current trajectory of digital twins reaches its limit.

The Missing Variable: Ground Behavior

At the center of this discussion is a variable that is consistently simplified, approximated, or treated as static: the ground itself.

In most infrastructure workflows, the ground is defined at the beginning of the project through site investigation, laboratory testing, and interpretation. A ground model is established, parameters are assigned, and variability is bounded within a range considered acceptable for design. From that point forward, the ground is treated as an input. It may be refined during construction, but it is not fundamentally redefined as part of the system.

This approach works as long as the ground behaves within the assumptions made at the design stage. It breaks down when construction begins to change the ground.

Excavation does not simply remove material. It redistributes stresses, alters confinement, and modifies the mechanical state of the remaining mass. Traffic not only applies the load. It changes density, rearranges particles, and can degrade structure over time. Repeated operations accumulate effects that are not linear and not always reversible. The material that was characterized at the beginning of the project is no longer the same after construction progresses.

In geotechnical terms, the response of the ground is controlled not only by its current stress condition, but by its stress history, fabric, and disturbance state. These evolve as construction advances. The system that needs to be understood is not a fixed profile with defined parameters. It is a state that is continuously changing in response to operations.

Current digital environments do not capture this well.

They can represent geometry as it changes. They can store measurements as they are collected. They can visualize deviations between predicted and observed behavior. What they do not do consistently is update the ground as a material system in a way that reflects its evolving mechanical state.

As a result, the connection between what is happening in the field and what is represented in the model remains incomplete.

This is often addressed through the observational method. Measurements are compared against expected behavior, and decisions are adjusted when thresholds are exceeded. This is effective, but it is reactive. It does not replace the need for a predictive understanding of how the ground will evolve under continued construction.

For terrestrial projects, this gap is managed through experience, redundancy, and the ability to intervene. Designs are adapted. Construction methods are modified. Additional investigation is performed if needed. The system tolerates a degree of uncertainty because there are mechanisms to respond to it.

For lunar and Martian construction, that tolerance is reduced.

Ground behavior will be influenced by conditions that do not have direct terrestrial equivalents. Low confinement affects how particles interact and how loads are transferred. Thermal cycling introduces changes in material response over time. The absence of water alters compaction and bonding mechanisms. Disturbance at the surface can have a disproportionate effect on performance because the stress regime is fundamentally different.

At the same time, the ability to adjust is limited. Equipment, energy, and operational flexibility are constrained. The cost of mischaracterizing ground behavior is not a delay in the schedule. It can be a loss of functionality. This is where the missing variable becomes critical.

The problem is not that digital twins lack data or computational capability. The problem is that the ground is not being treated as a dynamic system whose behavior evolves with construction and must be represented as such.

Why This Matters More in Space Than on Earth?

On Earth, the limitations described so far are real, but they are often manageable. Infrastructure projects are executed within a context that allows adjustment. Additional site investigation can be performed. Construction methods can be modified. Equipment can be mobilized or replaced. If the ground does not behave as expected, the system absorbs that uncertainty through intervention. That flexibility is built into how projects are delivered.

It is also supported by experience. Decades of tunneling, mining, and civil works have created a body of empirical knowledge that allows engineers to anticipate ranges of behavior, even when the ground is not fully characterized. The observational method works in this context because it is backed by the ability to respond.

That environment does not exist on the Moon or Mars.

The operational constraints are fundamentally different. Logistics are limited. Energy is constrained. Human intervention is reduced or delayed. Equipment must operate with a higher degree of autonomy and with far less tolerance for failure. The ability to “learn in the field” is restricted, not only by cost, but by feasibility.

At the same time, the ground conditions introduce additional layers of uncertainty.

For instance, near-surface regolith is governed by very low confinement. Particle interaction, load transfer, and deformation mechanisms do not follow the same patterns observed under terrestrial stress conditions. Thermal cycling is not a secondary effect; it becomes a dominant driver of material response over time. The absence of water removes mechanisms that are commonly relied upon in construction, from compaction to dust control to material binding. Surface disturbance, whether from traffic or landing operations, can alter the mechanical state of the ground in ways that propagate beyond the immediate area. These are not incremental differences. They change how the ground behaves as a system.

In this context, the assumption that construction can proceed with iterative adjustment becomes much weaker. If the ground response is not understood in advance, the options available during execution are limited. A misinterpretation of trafficability can affect mobility and logistics. An underestimation of deformation can compromise serviceability. An incorrect assumption about excavation effort can impact energy budgets and system performance. The consequences are not confined to local inefficiencies. They propagate through the entire system.

This is where the limitations of current digital approaches become more significant. A model that provides a coordinated view of assets and data, but does not adequately represent how the ground evolves, creates a false sense of confidence. It suggests that the system is understood because it is visible, even if the governing behavior is not fully captured.

On Earth, that gap can be closed during construction. In space, it must be addressed before and during operations with much tighter margins.

The problem is no longer how to improve coordination between models and data. It is how to ensure that the behavior of the ground, as it changes under construction, is sufficiently understood to support decisions that cannot be easily reversed.

The Core Gap: Data vs. Decision

At this point, the issue is no longer about tools or data availability. The industry has made substantial progress in both. Models are more advanced, sensor networks are more capable, and the ability to visualize and coordinate information across systems has improved significantly.

The gap sits elsewhere.

Data is being collected continuously. Measurements are available at a level of detail that was not possible before. Models can be updated, recalibrated, and refined as new information becomes available. From a technical standpoint, the system appears increasingly informed. But decision-making has not evolved at the same pace.

In practice, there is no consistent framework that translates evolving ground knowledge into clear thresholds for action. The question is not whether more data can be obtained. The question is whether the current level of understanding is sufficient to support a specific engineering decision with acceptable risk.

Comparison between conventional asset-centric digital twins and a ground-governed approach for infrastructure and planetary construction. The left side illustrates current practice, where models integrate geometry, systems, and sensor data but treat the ground as a static input. The right side presents a ground-governed twin, where the subsurface is modeled as a dynamic system evolving with excavation, loading, and disturbance, with uncertainty explicitly tracked. The bottom layer introduces Construction Readiness (CRL) as the decision framework linking ground understanding to design freeze, test validity, and operational acceptance.

During design, parameters are selected, and models are run based on available data and assumptions about variability. During construction, measurements are compared against predictions, and adjustments are made if deviations exceed predefined limits. This approach works, but it remains reactive. It does not provide a structured way to determine whether the underlying understanding of the system is adequate before committing to key steps.

As a result, critical decisions are often made in a zone of partial knowledge.

Designs are frozen based on interpretations that may not fully capture ground behavior. Systems are tested in environments that are assumed to be representative, without a rigorous basis for that assumption. Operations proceed with monitoring in place, but without a clear definition of what constitutes sufficient understanding of the ground response.

A system that is well instrumented and continuously updated creates the impression of control. Deviations can be detected. Trends can be observed. But detection is not the same as prediction, and observation is not the same as understanding.

This becomes particularly important when the system itself is evolving. If the ground is changing as a result of construction, then the reference against which decisions are made is also changing. Without a structured way to evaluate whether that evolving state is sufficiently understood, decisions remain dependent on interpretation rather than on defined readiness.

On Earth, this is often managed through experience and the ability to adjust. Engineers recognize when conditions are outside expected ranges and respond accordingly. The system relies on judgment, supported by data, but not fully governed by it. For lunar and Martian construction, that reliance becomes more critical and more exposed. There is a clear understanding that the question is no longer whether deviations can be detected and managed. It is whether the level of understanding of the ground system is sufficient before decisions are made that cannot be easily reversed.

The Missing Layer: Construction Readiness

Up to this point, the discussion has focused on what is available and where it falls short. The tools exist. The data exists. The models continue to improve. What remains unresolved is how this evolving knowledge is translated into decisions that carry real engineering consequences.

That translation is not currently formalized. Digital environments can show the state of the system. Models can simulate behavior. Monitoring can detect deviations. But none of these, on their own, define when the understanding of the ground is sufficient to move forward with confidence. The step from knowledge to decision is still largely dependent on interpretation.

This is where a construction readiness layer becomes necessary.

How the Construction Readiness Level (CRL) works upstream, complementing the Technology Readiness Level (TRL) and System Engineering (SE) (proposed by de Moraes, 2026).

It does not compete with Technology Readiness Levels, and it does not replace digital twins or model-based engineering. Each of those serves a different purpose. TRL evaluates the maturity of a system or technology. Digital environments provide a representation of the system and its evolving condition. What is missing is a structured way to determine whether the ground and construction system are sufficiently understood to support specific decisions.

It introduces the idea that readiness is not only a function of technology maturity, but also of how well the ground behavior and construction response are bounded. It shifts the focus from whether a system can operate to whether it can operate within a ground environment that is understood to an acceptable level.

At design freeze, the question is not only whether the system has been designed correctly, but whether the ground behavior has been sufficiently characterized to support that design without introducing uncontrolled variability. At test readiness, the issue is not only whether the system can be tested, but whether the environment in which it is tested is representative of the conditions it will encounter. At the start of operations, the decision is not only whether the system is functional, but whether the interaction between the system and the ground is understood well enough to avoid performance degradation.

Without this layer, these decisions rely on partial information. With it, they can be framed in terms of readiness. Not in a binary sense, but as a progression where the level of understanding of the ground and its response to construction is explicitly considered alongside the maturity of the system itself.

This does not require a new set of tools. It requires a different use of the information already being generated. Digital twins, monitoring systems, and models provide the inputs. Construction readiness defines how those inputs are evaluated in the context of decisions that cannot be easily reversed. It creates a link between evolving knowledge and action.

In that sense, it is not an additional framework layered on top of existing practice. It is the mechanism that connects them.

And in environments where the ground governs performance and the ability to adjust is limited, that connection becomes essential.

Toward a Ground-Governed Digital Twin

If digital twins are to support construction beyond coordination and visualization, their structure has to change. The shift is not incremental. It is conceptual.

The current approach starts from the asset and builds outward. Geometry is defined, systems are integrated, and data is layered on top to reflect performance. The ground enters this structure as an input. It is characterized, simplified, and then used to support the design.

For construction, and especially for off-planet environments, that sequence needs to be reversed.

The starting point has to be the ground as a system. Not as a profile or a set of parameters, but as a state that evolves under loading, excavation, and repeated operations. That state includes density, structure, stress condition, and disturbance, and it changes as construction progresses.

A ground-governed digital twin would treat that evolution as central, not peripheral. Machine interaction becomes one of the primary sources of information. In tunneling and raise boring, parameters such as torque, thrust, and penetration rate already reflect changes in ground response. In surface operations, mobility systems provide direct feedback through sinkage, slip, and energy demand. These are not secondary indicators. They are continuous measurements of how the ground is reacting to imposed loads.

In a ground-governed framework, these responses are not only recorded. They are used to update the state of the ground in a consistent way. The model does not simply store data. It evolves with it.

Uncertainty also changes the role. It is not treated as a range applied at the beginning of the project and refined as needed. It is tracked as part of the system, carried forward as the ground state evolves, and made explicit in how decisions are evaluated. The question is no longer whether the model matches observations at a given point. It is how confident we are in the current representation of the ground and how that confidence affects the next step.

The construction sequence becomes part of the model. Excavation, placement, traffic, and loading are not external processes applied to a predefined system. They are the mechanisms through which the system changes. The twin has to reflect that interaction step by step.

This is not a matter of increasing model fidelity or adding more sensors. It is a matter of coupling behavior with representation.

For lunar construction, the ground will govern trafficability, load transfer, excavation effort, and long-term performance. If its evolution is not captured, the model will remain descriptive rather than predictive.

The objective is not to create a perfect replica of the physical system. That is not achievable. The objective is to create a representation that evolves in a way that is consistent with the mechanics of the ground and that makes uncertainty visible in the context of decisions.

Beyond Space

The argument does not stop at lunar or Martian construction. In many ways, space simply exposes a limitation that already exists on Earth but is currently managed rather than resolved.

Terrestrial projects operate with a level of flexibility that masks the gap. When ground behavior is not fully understood, adjustments are made during construction. Methods are modified, additional support is installed, sequences are adapted, and monitoring is intensified. The system works because it can absorb uncertainty through intervention.

But that comes at a cost.

Delays, redesign, inefficiencies, and conservative assumptions are often the consequence of operating without a fully integrated understanding of how the ground evolves during construction. These are accepted as part of the process, not because they are optimal, but because they are manageable. As construction moves toward higher levels of automation, that tolerance begins to erode.

Autonomous and semi-autonomous systems require a different level of predictability. Earthworks equipment, tunneling systems, and mining operations that rely on reduced human intervention need a more consistent understanding of ground response. The ability to adjust in real time still exists, but it must be embedded within the system itself rather than relying on external decision-making.

And more importantly, this is where the same gap becomes more visible.

If the ground is not represented as a dynamic system, autonomous operations are forced to operate conservatively. Productivity is reduced to maintain safety margins. Energy consumption increases. Equipment performance is limited by uncertainty rather than by capability.

In mining, this appears in the form of inefficiencies in material handling and excavation. In tunneling, it affects advance rates, cutter wear, and support requirements. In infrastructure, it influences settlement control, serviceability, and long-term performance.

The tools to address this are already partially in place. Machine data, monitoring systems, and numerical models exist. What is missing is the integration of these elements into a framework that treats ground behavior as central and links it directly to operational decisions.

The same applies to maintenance and lifecycle management. Infrastructure performance over time is often governed by how the ground continues to respond to loading, environmental conditions, and aging. Without a dynamic representation of that behavior, maintenance remains reactive. Issues are identified after they develop, rather than being anticipated as part of the system’s evolution.

The relevance of space construction is that it removes the ability to rely on adjustment as a primary strategy.

What must be done off planet becomes what should be done on Earth.

By forcing a more explicit treatment of ground behavior, uncertainty, and decision making, lunar and Martian construction provides a framework that can be translated back into terrestrial practice. The same principles that enable reliable operations in constrained environments can improve efficiency, resilience, and performance in environments where flexibility still exists.

Practical Transition Path from Data to Decision

If the objective is to move from current digital environments toward a system that can support construction decisions on the Moon, the path does not start from scratch. The components already exist across multiple industries. The challenge is how they are combined and what role they play.

The starting point is already visible in practice. In tunneling and raise boring, machine interaction with the ground provides continuous feedback. Torque, thrust, penetration rate, and drilling response are direct expressions of how the material is behaving at that moment. In oil and gas, real-time data streams are used to update subsurface interpretations during drilling, allowing decisions to be made within the operational loop. On the Moon, rover interaction, mobility resistance, sinkage, and energy demand play the same role. These are not indirect observations. They are measurements of ground response under load.

The next step is to move from interpretation to representation.

Instead of treating these signals as indicators to be reviewed, they need to be used to define the state of the ground as it evolves. Not as fixed parameters established at the beginning of the project, but as a condition that is updated continuously based on interaction. Density, structure, disturbance, and load response become part of a living system, not a predefined model.

At that point, uncertainty can no longer be treated as an initial assumption that is gradually reduced. It becomes part of the system itself. It is tracked as the ground evolves, carried forward with each update, and made explicit in how the system is understood at any given time. The question is not whether uncertainty exists, but how it affects the next decision.

The evolving understanding of the ground has to be connected to decision gates. Design freeze is no longer based only on completing the design. It depends on whether ground behavior has been sufficiently bounded to support that design. Testing is no longer defined only by system readiness. It depends on whether the test environment represents the ground conditions that will govern performance. Operations are not only a matter of system functionality. They depend on whether the interaction between the system and the ground is understood well enough to manage performance over time.

It does not require abandoning existing tools or frameworks. It requires reordering them. Machine data, monitoring, and modeling already provide the inputs. What changes is how those inputs are used to define the ground as a dynamic system and how that definition is tied to decisions that carry consequences.

How Earth-Based Mining, Oil & Gas, and Infrastructure Projects Can Cover the Gaps?

Digital twins will continue to evolve. The integration of models, data, and systems will improve, and the ability to represent infrastructure in a coordinated digital environment will become more refined.

What remains unresolved is what those representations actually mean when decisions have to be made.

In infrastructure, and particularly in construction, performance is not governed by the model. It is governed by the ground. If the behavior of the ground is not sufficiently understood, no level of model integration will compensate for that gap. The system may appear controlled, but the governing response remains uncertain.

For lunar and Martian construction, the solution is not to build more complex digital environments. It is to structure them around ground behavior and decision-making from the beginning.

Ground characterization must transition from a one-time input to a continuously updated state. Penetration resistance, mobility response, excavation effort, and machine interaction need to be treated as primary data streams, not secondary observations. On the Moon, this means integrating rover performance, penetration tools, and construction equipment feedback into a unified interpretation of regolith state.

Construction has to be part of the model. Excavation, traffic, and repeated loading are not external actions applied to the system. They are the mechanisms that define how the ground evolves. Any digital representation that does not include that interaction will remain descriptive rather than predictive.

Uncertainty must be carried forward, not reduced to a range at the beginning of the project. Decisions should be made with explicit awareness of how well the ground is understood, not based on assumed convergence.

And most importantly, decisions must be tied to readiness. Design freeze, testing, and operations need to be gated not only by system maturity, but by how well the ground response has been bounded.

This is where current terrestrial industries provide a path forward.

Mining already operates with continuous feedback between excavation and material response. The ability to track movement, production, and local variability can be extended beyond geometry into ground state interpretation.

The oil and gas industry has demonstrated that subsurface models can evolve during operations. Real-time data integration, combined with decision-making in the loop, provides a direct analogue for how construction systems could adapt to changing ground conditions.

Tunneling and raise boring bring the closest link between the machine and the ground. Torque, thrust, penetration rate, and drilling response are direct expressions of ground behavior. These are not proxies to be reviewed after the fact. They can form the basis of a continuously updated understanding of the material being excavated.

None of these industries has fully closed the loop. But together, they define the components of a system that can.

For space construction, the opportunity is to integrate these approaches from the outset. Not as separate workflows, but as a single framework where ground state, construction response, and decision making are directly connected.

That integration is what enables construction readiness.

It ensures that digital representations are not only accurate in form but meaningful in function. It provides a basis for decisions that cannot rely on adjustment after the fact. And it aligns the model with the system that ultimately governs performance.

Digital twins will not fail. They will continue to add value in coordination, visualization, and data integration. But they will remain incomplete until they answer a more fundamental question. Not how well the system is represented. But whether the ground that governs it is understood well enough to build.

For the Moon and Mars, that question defines success.

And the answer will not come from models alone. It will come from how we integrate ground behavior, construction, and decision making into a single, defensible framework.

Roberto de Moraes

Author | SpaceGeotech Founder