Integrating life support systems into 3D-printed extra-terrestrial habitat envelopes

Abstract

When designing and constructing habitats on Mars, one of the main challenges is the extreme environment, which lacks a breathable atmosphere and atmospheric pressure, and is characterized by extreme temperature fluctuations and high radiation levels. The habitat is situated within a lava tube, providing natural shielding from radiation, micrometeoroids, and extreme temperature fluctuations. This challenge is addressed in TU Delft’s Rhizome 2.0 project by integrating a Life Support System (LSS) into the envelope of a Martian habitat to maintain the necessary environmental conditions for human survival. The research explores the integration of the LSS into the approximately 1.5-m-thick building envelope that is constructed from prefabricated 3D-printed Voronoi-based components that interlock and are assembled on-site by a swarm of rovers and human assistance where needed. The integration is implemented by structurally analyzing the envelope to generate a point cloud of stress data. Based on the stress concentrations and LSS conduit path lengths, a single objective optimizer determines the shortest routes for the LSS conduits while avoiding high stress concentrations. This results in a structurally optimized building envelope with strategically incorporated branching negative spaces that accommodate the LSS infrastructure.

1 Introduction

Constructing extraterrestrial habitats requires consideration of the extreme environment, which lacks a breathable atmosphere and atmospheric pressure, and is characterized by extreme temperature fluctuations and high radiation levels. Various precedent case studies in the last decade aim to address these challenges, for instance, The Mars Icehouse and The ICE Home, using ice as the main construction material, as it is more effective against radiation than regolith-based constructions. They require, however, extra measures to keep the ice from sublimating into the atmosphere, as the use of inflatable plastics. Regolith-based constructions such as The Mars Habitat, relying on sintering regolith as the main construction material, and X-House or MARSHA involving 3D printing using basalt composites are monolithic (inter al. Bier et al., 2021), while the TU Delft’s Rhizome project involves assembly of 3D printed prefab components.

Intense radiation, micrometeoroids, extreme temperatures, and the absence of a pressurized atmosphere make survival of humans on Mars and the Moon only possible by deploying robust protective systems such as a radiation-resistant building envelope with integrated Life Support Systems (LSS). The LSS comprises physico-chemical systems that actively and/or passively regulate the environment within an extraterrestrial habitat. Its primary functions are the management of atmosphere, water, and waste, each supported by dedicated subsystems. The atmosphere management subsystem draws used atmosphere from the habitat, filters and recycles it to remove particulate matter, scrubs excess CO2, replenishes oxygen, recovers moisture, and removes trace contaminants. The water management subsystem collects wastewater and filters it to achieve the required level of purity. The purified water is then stored and redistributed to all locations within the habitat that require clean water. The waste management subsystem handles trash, human, and process-generated waste.

An example of an LSS that consists of several subsystems is the Micro-Ecological Life Support System Alternative (MELiSSA), developed by the European Space Agency (ESA). This is a regenerative life support system for long-term human space missions (inter al. Hendrickx et al., 2006). The integration of such systems in the ESA co-funded Rhizome 2.0 project involved the development of a Martian habitat design relying on the robotic assembly of regolith-based 3D printed building components designed to interlock (Bier et al., 2021). The components are 3D printed within a temporary pressurized production environment, allowing the atmospheric conditions during fabrication to be controlled.

Component lengths range from 500 mm to 1,500 mm, with an approximate thickness ranging from 250 mm to 500 mm. Components at the lower end of this range can be handled and placed by a single rover, while larger components require cooperative transport and installation by two rovers operating in a coordinated lifting and placement configuration. The interlocking relies on a Voronoi-based optimization involving Voronoi cell densities increasing in high-stress regions and decreasing in lower-stress areas. The advantage of this approach over traditional topology optimization approaches is the maintenance of a consistent minimum envelope thickness of 1.5m, which is a critical constraint for the assembly process. The Voronoi-based approach utilizes cell orientation and sizing to create a stiff, materially efficient shell, which ensures that the top surface remains navigable for the autonomous swarm of rovers to assemble the habitat via an integrated ramp system (Aslaminezhad et al., 2024). The building skin overhang angles are constrained to 45° relative to the foundation to avoid the need for support structures during assembly.

The assembly of the components is carried out by a swarm of rovers working collaboratively with human operators, when necessary, with robots handling automated, less complex assembly tasks and heavy lifting, while humans intervene in situations requiring cognitive skills beyond the capabilities of the current AI systems (Bier et al., 2025)

2 State-of-the-art

There are two main approaches to integrating LSS in the habitat. The first is on the surface of the envelope (Figure 1 left) and the second inside the building envelope (Figure 1 right). Both have advantages and disadvantages (Bier et al., 2021), but considering the negative psychological impact of cluttered indoor environments, integration in the building envelope is preferred (inter al. Obeidat and Obeidat, 2024). Rhizome 2.0 integrates LSS systems into the building envelope itself, thus aiming at improving psychological comfort. The underlying assumption is that minimizing exposed conduits on interior surfaces results in a less cluttered visual environment, which contributes to improved psychological comfort. Adapting the geometry to host cavities for building services is possible in 3D printing due to the layer-by-layer nature of the process (Meisel et al., 2022). There are various examples of multi-functional building components, for instance, Taher et al. (2023) showcase a wall design with an integrated displacement ventilation system. Another example of functionally integrated 3D-printed curtain wall components consisting of a reticulated network of channels connecting the interior with the exterior boundaries of the wall incorporated electrical and water utilities (Goidea et al., 2024). Jipa et al. (2019) demonstrated a structurally optimized concrete slab featuring profiled soffits and hollow spaces that can integrate ducting for heating, ventilation, and cooling, which are distributed through funicular-shaped cavities within the slab. The integration presented in this paper is for a fragment of a building envelope with cavities that range from small (50 mm) for electrical to large, up to room-sized, technical spaces.

FIGURE 1

For the integration of LSS, existing life support programs, including NASA’s Advanced Life Support, ESA’s MELiSSA, and Russia’s BIOS-3 experiments, provide important foundational knowledge. However, these efforts largely investigate isolated bioregenerative or physicochemical subsystems rather than integrated, mission-ready life support architectures (Liu et al., 2021; Häder et al., 2018; Salisbury et al., 1997). The LSS in the Rhizome project is considered part of a tightly coupled ecological–technological network (Raihan, 2026) and integrated in the initial habitat layout in order to prioritize efficiency, accessibility, maintainability, and crew safety (Cohen et al., 2024). Cohen et al. note that decentralization of the life support system increases redundancy, improves spatial efficiency, and thereby enhances crew survival probability. Amini et al. (2022) proposed a conventional service distribution strategy in which the water supply is routed through the floor construction, while air-related systems (including ventilation, humidity control, and fresh air supply) are integrated into a suspended ceiling system. Wastewater lines are likewise accommodated within the floor assembly. However, this configuration is presented at a conceptual level only and has not been developed into a detailed engineering design. Current limitations requiring further investigation include the lack of validation of many systems under lunar (0.16 g) and Martian (0.38 g) gravity conditions. As a result, system performance is often extrapolated from microgravity experiments or Earth-gravity analogues (Raihan, 2026).

3 Contribution

The presented case study integrates the LSS within a building envelope employing a single-objective optimization algorithm that minimizes conduit lengths while avoiding high-stress concentrations. Stress values and path lengths are combined into a unified scoring function that penalizes suboptimal solutions. The resulting paths can subsequently be modeled as negative spaces in the building skin that can host the LSS conduits. This design approach is compatible with the constraints for robotic fabrication and assembly, as well as maintenance and accessibility.

4 Methodology

The LSS consists of equipment and conduits that are routed through the building envelope. These LSS conduits can be integrated into the building skin by generating branching negative spaces of variable sizes, thereby creating space for equipment, cables, wires, and piping. These negative spaces are scalable and accessible by robots and/or humans for maintenance. Conduit dimensions range from a minimum of 50 mm to accommodate access, sleeve/harness routing, and service integration, up to room-scale voids that house centralized technical systems and provide maintenance and emergency access. Access to LSS/ECLSS systems is critical in Martian habitat environments. In the event of failures, crew members must be able to rapidly access technical spaces and service conduits. Locally, wall thickness is then increased to integrate these expanded service volumes. The primary spaces in the habitat are defined as the dedicated areas housing the technical equipment required for life support systems, including components such as pumps, filters, and reaction chambers. These systems are responsible for functions such as CO2 scrubbing, atmosphere filtration, water purification, and wastewater separation in extraterrestrial habitats. Generating the cavities to integrate the LSS conduits will be iteratively assessed with a Finite Element Analysis (FEA) model to analyse the structural performance of a fragment of the envelope and identify optimal locations in areas of minimal stress. The FEA model will generate a point cloud with principal stress data that identifies regions of relatively high and low structural stress. Together with the functional requirements of the LSS, these data form the basis for routing optimization. This iterative approach will optimize the LSS cavities in the building envelope.

4.1 Structural analysis and LSS path finding

For this proof-of-concept study, we examine three fragments of the overall building envelope to develop an automated path adjustment workflow for the LSS cavities, so that it avoids high stress concentrations. The initial routing path is represented by a simplified continuous curve running through the fragment. This simplified curve represents the connection between a technical space and a service outlet for utilities such as ventilation, electrical power, and other building services. The automated form-finding algorithm is intended to optimise its placement to reduce its impact on the structural performance of the fragment. For this, an FEA model setup of the Martian habitat design was implemented in Karamba, a Grasshopper plugin, which calculated stress densities and lines based on the self-weight of the structure (Figure 2). Fixed supporting nodes were applied to the points located close to the bottom of the building geometry to simulate simplified foundations. The fragment was modelled as a BREP and subsequently meshed into 22780 vertices and 22949 faces. Since the mesh is made from a BREP definition, it can be meshed in any resolution. The current resolution was found to be an acceptable balance between resolution and calculation time (3 min). When the shell was added to Karamba’s shell-based structural analysis, it allowed for the specification of constant shell thickness, and principal stress values were extracted from the model at 10 predefined offsets through the shell thickness, which included the outermost face, innermost face, and the equally spaced thickness offsets. The principal stress values resulted from this, together with their respective point positions in the geometry, resulting in a point cloud of stress data. The 3D point cloud represents the building envelope, where each point contains stress concentration data. This analysis revealed variations in stress levels over the geometry.

FIGURE 2

After this analysis, three smaller-scale fragments were extracted to test the methodology. The routing paths are then optimized using the Galapagos single-objective evolutionary solver within Grasshopper (Rutten, 2013). The optimization seeks to relocate conduit paths away from structurally critical regions while maintaining feasible routing trajectories. To achieve this, selected curve control points are displaced within local two-dimensional planes oriented normal to the routing curve (Figure 3). Candidate positions are constrained to remain within the wall fragment geometry, ensuring that only valid routing configurations are evaluated. Each control point is associated with two independent design variables corresponding to horizontal and vertical displacement within its local optimization plane. These parameters define the search space explored by the evolutionary solver. The start and end points of each routing curve path remain fixed throughout the optimization process, preserving connectivity between adjacent building fragments. The evolutionary solver (Galapagos) was configured with 50 iterations and a maximum stagnation limit of 50 generations. A population size of 200 individuals was used, with an initial boost factor of 12 to increase early exploration of the solution space. During evolution, 40% of the population was maintained between generations, while an inbreeding factor of +75% was applied to bias local exploitation around high-performing solutions.

FIGURE 3

The stress field obtained from the global structural analysis is normalized to a range between 0 and 10 to provide a consistent fitness landscape for the optimization process. During each iteration, the modified control points on the 2D planes are reconstructed into a NURBS curve and sampled at 100 mm intervals. For each sample point, the nearest stress value is retrieved from the stress point cloud. The average stress along the reconstructed curve is subsequently calculated and used as the primary optimization objective.

To discourage unnecessary detours, a secondary path-length penalty is added into the fitness function. The path length is normalized and weighted such that it contributes approximately 10% of the total fitness score. This weighting provides a preference for shorter routes while preserving the primary objective of minimizing exposure to regions of elevated structural stress. To evaluate performance, the optimized curves are compared with the original curves and shortest path solutions optimized only for length. These shortest path curves have also been calculated by Galapagos, where the fitness function was changed to the total length of the curve. The results of this optimization and the differences in length are listed in Table 1. The latter serves as a geometric baseline without structural considerations, while the proposed method includes stress-informed optimization. Table 1 also shows the performance comparison in terms of path length and stress concentrations over the length of the curve.

5 Results and discussions

The optimization process iteratively evaluated alternative routing configurations and identified solutions that minimize the combined fitness score. Across the analyzed fragments, the optimized routing paths consistently shifted toward regions of lower structural stress, resulting in reductions of approximately 30%–43% in average stress exposure compared to the initial routing configurations, at the cost of extra path length (17%–61.6%) (Table 1). Figure 3 shows fragment 1 and the corresponding path optimization process. In all cases, the solver demonstrates clear convergence. The stress points, with relative stress lower than 4.5, are displayed. In Figure 3, it is visible that the optimized curve consistently aligns with the lower-stress regions while simultaneously minimizing path lengths.

The optimization of negative branching spaces within the habitat facilitates structural integration of the LSS conduits in the building skin while avoiding high-stress areas, this presents a novel and automated framework to conserve resource utilization in extreme conditions. The proof-of-concept demonstration on the fragment, however, presents several limitations that will be addressed in future work. For example, the use of simplified structural models and a single-objective formulation with fixed weighting. Fixed weights should be improved by including a larger number of fragments and eventually at the entire habitat level needs to be optimized and investigated, with possibility for LSS cavities to generate off-shoot branches. Moreover, material properties were approximated using predefined concrete parameters in the Karamba plugin, and loads considered in FEA being limited to self-weight and approximate boundary conditions. The loading conditions, assigned via gravity loads along the envelope, were rough approximations of actual conditions solely for the purpose of generating principal stress results. Future work applying this framework should consider additional stresses induced by temperature fluctuations, wind, seismic activity, internal atmospheric pressure, and variable loads from internal use. Moreover, foundation conditions should be modelled more accurately to the exact expected conditions of the structure and soil characteristics. Nevertheless, the demonstration on the fragment presents an opportunity for automated LSS cavities integration within the building skin that prioritises structural performance, and thereby requires less material resources, enabling optimal placement of LSS in relation to the subdivision of the envelope into optimized,

The optimization results demonstrate convergence, suggesting that the proposed formulation is well-posed and yields feasible solutions. However, Galapagos is a stochastic evolutionary solver and cannot guarantee global optimality. The optimizer reduces stress along the routing curve while controlling path length. In the current formulation, the path-length penalty contributes 10% of the fitness score, and thus has a limited influence on the resulting solutions. Increasing this weighting would produce shorter paths at the expense of higher stress exposure. Future improvements will focus on refining the balance between distance and stress penalties, for example, by increasing the weight of shorter paths or introducing stronger penalties for high-stress regions above defined thresholds. Higher-resolution evaluations are expected to further improve solution accuracy.

After the FEA output for the conduit path, additional constraints should be checked to ensure compliance with critical requirements for the design’s constructability of the negative spaces within the facade’s components. For example checking if the 3D-printable components and ensuring that the void fraction of the total wall volume is limited to 10%–35%. Another check should ensure that equipment can be integrated and that humans and robots can access it easily. This is done by considering the maximum allowable slope angles. The slopes require inclinations that facilitate human movement and enable automated, robotized inspections and maintenance of the LSS infrastructure. Inspection and access channels should be dimensioned to a minimum of 800 mm for human entry and 300 mm for robotic access. After the negative spaces have been placed inside the building envelope, the structural analysis should be reiterated to determine the final componential logic and density of the structural infill. This ensures that the changes in the structure are compensated for. Future work will also include the development of a full-scale (1:1) prototype of a building skin fragment, comprised of physically 3D-printed components, to test the interlocking logic and LSS integration. Figure 4 already presents a digital schematic habitat design, including selected extracted components from a representative fragment and an exploded view illustrating the integrated negative spaces. The design methodology leverages the inherent properties of the Voronoi-based components. Their geometry allows individual components to settle into configuration during assembly, enabling minor positional deviations to be accommodated while maintaining overall structural stability. The results of the 1:1 scale experiments will be used to quantify construction tolerances and to determine how these tolerances affect the minimum envelope thickness requirement of 1.5 m. The effects of the Martian environment on the material properties and long-term performance of the prefabricated 3D-printed components after fabrication will be investigated in future work. Currently, research integrating AI-supported illumination for improving human-comfort is being implemented and future research will include human factors assessment of the full-scale design with integrated lighting. Various systems require further consideration as for instance the integration of ECLSS and other subsystems such as power generation and supply, communication, etc., and will be considered in future steps.

FIGURE 4

6 Conclusion

This study presented a stress-informed routing methodology for Life Support System (LSS) conduits within structural building envelopes. The approach integrates structural analysis with geometric path optimization, using a spiral-based initial routing strategy and a single-objective evolutionary solver (Galapagos) to minimize stress exposure while maintaining feasible conduit paths. Across three representative building fragments, the proposed method consistently improved routing performance compared to both the initial spiral configuration and a shortest-path baseline. The optimized solutions reduced average stress exposure by approximately 30%–43%, with Fragment 2 achieving the highest reduction (43.4%). These improvements were achieved at the cost of increased path length, reflecting a controlled trade-off between structural performance and routing efficiency. The results demonstrate that incorporating structural stress fields into routing optimization can effectively guide LSS pathways toward lower-stress regions within the building envelope. Furthermore, the method shows consistent behavior across different geometric fragments, indicating the robustness of the proposed framework. Limitations of the current study include the use of simplified structural models, a single-objective formulation with fixed weighting, and evaluation on a limited number of fragments. Future work should extend the framework to full-scale building models, investigate multi-objective optimization strategies, and incorporate branching network topologies for complete LSS system design.

References

• 1

  AminiK.MoradiM.TeymooriP.VossoughiB.JanabadiE.FayazR. (2022). Design of a set of habitat units and the corresponding surrounding cluster for long-term scientific missions in the pre-terraforming era on Mars. Icarus385, 115119. 10.1016/j.icarus.2022.115119

  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  AslaminezhadA.HiddingA.BierH.CalabreseG. (2024). In-Situ vs. prefab 3D printing considerations for CO2-free Pop-up architecture. SPOOL11 (1), 71–80. 10.47982/spool.2024.1.05

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  BierH.VermeerE.HiddingA.JaniK. (2021). Design-to-Robotic-Production of underground habitats on Mars. Spool8 (2), 31–38. 10.7480/spool.2021.2.6075

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  BierH. H.HiddingA. J.PrendergastM.PeternelL. (2025). “Computer Vision- and human-robot interaction-supported assembly for collaborative off-earth habitat construction,” in Intelligent Computing and Automation. FICTA 2024 2024. Editors BhatejaV.DeyM.SimicM. (Singapore: Springer), 421. 10.1007/978-981-96-0143-1_9

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  CohenM.MettsJ.BarkerD.DesjeanM.HastingsD.WeitzL.et al (2024). ECLSS -First Space Habitat Architecture” (ICES-2024-193) ECLSS-first Space Habitat Architecture.

  • Google Scholar
  • View reference in article

• 6

  GoideaA.PopescuM.JohanssonA. T.Andre’enD. (2024). Algorithmic modeling of functionally graded meta-materials in 3D printed building envelopes. Archit. Intell.3 (1), 25. 10.1007/s44223-024-00068-x

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  HäderD. P.BraunM.HemmersbachR. (2018). “Bioregenerative life support systems in space research,” in Gravitational Biology I: Gravity Sensing and Graviorientation in Microorganisms and Plants. Cham: Springer International Publishing, 113–122.

  • Google Scholar
  • View reference in article

• 8

  HendrickxL.De WeverH.HermansV.MastroleoF.MorinN.WilmotteA.et al (2006). Microbial ecology of the closed artificial ecosystem MELiSSA (Micro-Ecological life support system alternative): reinventing and compartmentalizing the Earth’s food and oxygen regeneration system for long-haul space exploration missions. Res. Microbiol.157 (1), 77–86. 10.1016/j.resmic.2005.06.014

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  JipaA.BarentinC. C.LydonG.RippmannM.ChousouG.LomaglioM.et al (2019). 3D-printed formwork for integrated funicular concrete slabs. Proc. IASS Annu. Symposia2019 (6), 1–8. 10.3929/ethz-b-000387460

  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  LiuH.YaoZ.FuY.FengJ. (2021). Review of research into bioregenerative life support system(s) which can support humans living in space. Life Sci. Space Res.31, 113–120. 10.1016/j.lssr.2021.09.003

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  MeiselN. A.WatsonN.Bile’nS. G.DuarteJ. P.NazarianS. (2022). Design and system considerations for construction-scale concrete additive manufacturing in remote environments via robotic arm deposition. 3D Print. Addit. Manuf.9 (1), 35–45. 10.1089/3dp.2020.0335

  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  ObeidatA.ObeidatA. (2024). Interior design strategies for improving quality of life: how can residential spaces reflect a healthy lifestyle and psychological comfort. Int. J. Hous. Sci. Appl. 45, 71–86. 10.70517/ijhsa4547

  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  RaihanA. (2026). Toward sustainable living in space: a review of environmental control and life support system technologies. Space Habitat.2 (1), 100047. ISSN 2950-6166. 10.1016/j.spaceh.2025.100047

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  RuttenD. (2013). “Galapagos: on the logic and limitations of generic solvers,” Galapagos Log. Limitations Generic Solvers, 83. 132–135. 10.1002/ad.1568

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  SalisburyF. B.GitelsonJ. I.LisovskyG. M. (1997). Bios-3: siberian experiments in bioregenerative life support. Bioscience47 (9), 575–585.

  • Pubmed Abstract
  • Google Scholar
  • View reference in article

• 16

  TaherA.AssutS.van der SpoelW. (2023). An integrated workflow for designing and fabricating multifunctional building components through additive manufacturing with clay. Buildings13 (11), 2676. 10.3390/buildings13112676

  • CrossRef
  • Google Scholar
  • View reference in article