AIAA SciTech Forum 10.2514/6.

2024-0768
8-12 January 2024, Orlando, FL
AIAA SCITECH 2024 Forum

Creation of a Structural Model to Facilitate an OpenMDAO

Based Lunar Rover Parametric Sizing Tool

Alberto Cardenas-Melgar∗ , Neel Puri † , Bradford Robertson ‡ , and Dimitri Mavris§

Georgia Institute of Technology, Atlanta, GA, 30308

The Artemis mission aims to explore more of the lunar surface than ever before. Lunar
rovers can aide this goal, yet the rover design space is vast and complex. Rover designers can
benefit greatly from understanding the impact of high level design choices on the final design.
A parametric rover sizing tool was created to fulfill this need. This paper outlines the tool’s
framework and details how the rovers structural components are modeled and sized. The three
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

structural models – the pressure vessel, chassis, and suspension – are evaluated against CAD
models and published masses of conceptualized rovers. The validation showcases a 3% error
for the pressure vessel, 15% for the chassis, and 2% for the suspension system between the
respective models and published values. An example design space analysis is conducted and
discussed to highlight the developed capability and it’s relevance to the field.

I. Introduction
he Artemis missions mark a new era of lunar exploration, with the goal of returning humans to the Moon. The
T mission faces significant challenges in building on the legacy of the Apollo missions while also developing new
tools and technologies for lunar exploration [1]. One such tool is a pressurized rover, which can transport astronauts
across the lunar surface in a safe and comfortable living environment for extended missions [2].
For years, there have been many innovative lunar rover concepts ranging across senior design projects [2, 3],
corporate sponsored concepts [4, 5], academic concept explorations [6], contracted prototypes like JAXA’s Toyota rover
[7], to built and tested vehicles like NASA’s Lunar Electric Rover (LER) [8, 9]. A large percentage of these concepts,
however, focus on a single set of requirements and thus produce singular design-points. For the Artemis missions,
understanding the final design’s sensitivity to its requirements can prevent many costly mistakes. A parametric sizing
tool can provide this ability by showing engineers how changes in certain top level requirements trickle down to the final
rover design.
The formulation of such methods has to focus on using first principles and core assumptions to create models that
can integrate well into a multi-disciplinary tool. Once integrated, however, the tool would be able to highlight patterns
in the design single-discipline engineers might overlook. For example, how the length of the design mission interacts
with the number of wheels to limit the maximum number of support beams the chassis can have.
This paper outlines a framework for such a parametric rover sizing tool, and focuses on the modeling of key structural
components. The structural requirements are handled by three components, the pressure vessel which houses the
astronauts, the chassis which connects the vessel to the suspension and provides rigidity, and the suspension which
connects the chassis to the wheels. While each component has individually been understood a great deal for things like
automobiles and submarines [10–12], a combination of them that has to withstand the harsh lunar environment for days
on end is a realm relatively untouched. Previous studies have explored these combination concepts in depth, such as [2]
and [8], and those studies are used as validation for the results of the models outlined here. The resulting models are
used in an example design space study to showcase how they, and the larger tool, can be useful for rover designers.

II. Methodology
To begin development a rovers core functions had to be defined. The resulting functions were then organized based
on their inter-dependencies into an N2 diagram, a simplified version is shown in Fig. 1. This N2 diagram formed the
basis for the tool. Each of the elements shown in Fig. 1 represent a different physics-based model that was created. This
∗ Graduate Research Assistant, Aerospace Systems Design Laboratory, School of Aerospace Engineering. AIAA Student Member.
† Graduate Research Assistant, Aerospace Systems Design Laboratory, School of Aerospace Engineering. AIAA Student Member.
‡ Research Engineer II, Aerospace Systems Design Laboratory, School of Aerospace Engineering. AIAA Member
§ S.P. Langley Distinguished Regents Professor, Aerospace Systems Design Laboratory, School of Aerospace Engineering. Fellow AIAA.

Copyright © 2024 by Alberto Cardenas-Melgar, Neel Puri, Bradford Robertson, Dimitri Mavris.
Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.
 paper outlines the structural models (Pressure Vessel, Suspension, and Chassis), and [13] outlines the thermal, power
and power management and distribution (PMAD) models.

C&DH
Navigation
Science

Furnishings
Ingress/Egress

Pressure
Vessel
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Key ECLSS
Inputs Thermal
Volume
Power &
Thermal Load PMAD
Power
Movement
Mass & Chassis

Converged Rover Sizing

Fig. 1 Diagram showing an N2 diagram of the core rover functions.

Beginning in the top left, the Communication, Data Handling, Navigation, and Science payloads are mostly invariant
to mission length or crew size. Hence, they can be defined as rubberized payloads. Next, the Furnishings and
Ingress/Egress model changes with crew size but not mission length. Both of these sets of models require definition but
no-feedback from other models and are hence at the top of the diagram. Next the Pressure Vessel model, described
in Section II.A, takes in the volume requirements of all the other systems and sizes the vessel appropriately. The
Environmental Control and Life Support System (ECLSS) & Thermal models were coupled because they were originally
modeled using NASA Johnson Space Center’s Advanced Life Support Sizing Analysis Tool (ALSSAT) which couples
the two together [14]. The thermal model was later decoupled from ECLSS and ALSSAT, and the resulting model is
what is outlined in [13]. Despite this, the sizing tool takes thermal loads and uses that, along with mission parameters,
to size the ECLSS and thermal systems. Similarly all the power requirements of all the systems are combined to size the
Power and PMAD systems. Finally, the mass of every system is used to size motors, wheels, suspension and the chassis.
Once every model is sized, the tool follows the feedback loops to continually resize all the components until it converges
on a solution. The resulting solution defines key aspects of a lunar rover sized for the given set of input parameters.
This tool can be used to understand how lunar rover design responds to varying mission requirements or to conduct
design trades (such as the analysis done in [13]). It uses the OpenMDAO framework (Open-source Multidisciplinary
Analysis and Optimization [15]). Having defined the larger context for the models, the proceeding sections outline the
structural components were modeled.

A. Pressure Vessel
The pressure vessel, or frame, plays an essential role in creating structural support, protecting vital components,
shielding against space radiation and lunar dust, and pressurizing the system. To ensure a parametric capability, the
model’s key inputs are geometrical parameters such as the length, width, and height of the vehicle. Several assumptions
were made to achieve this including a monocoque design, defeatured pressure vessel, and a rectangular shape with
rounded corners for the cross-section of the pressure vessel, see Fig. 2 as a reference. This capability simplifies the

2
 complexity associated with pressure hull design but still provides a reasonable representation.
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Fig. 2 Example of the pressure vessel geometry with Fig. 3 Section of the pressure vessel’s cross-section
parameterized rounded corners. showing the different layers of material

The model has seven design inputs, as well as an additional input from the ECLSS system, to determine the geometry
and composition of the pressure vessel. The inputs are the length, height, and width of the vessel, a corner curvature
ratio (to vary between a rectangular and circular cross section), the number of layers and their corresponding insulation
layers, the density of those layers (see Fig. 3) as a reference), the yield strength of outer layer’s material (assuming it is a
ductile material), and the internal pressure for the failure analysis of the pressure vessel.
The volume of each layer is calculated in two steps. First, the volume of the vessel is calculated for the longitudinal
section, followed by the computation of the caps’ volume. The volume of the vessel (𝑉 𝑜𝑙 𝑝𝑣 ) is obtained by multiplying
the cross-sectional area (𝐴𝑐𝑠 ) by the corrected length (length of the vessel minus twice the radius, 𝐿 𝑐𝑜𝑟𝑟 𝑒𝑐𝑡𝑒𝑑 ), see
Eqs. (1) & (2).
𝑉 𝑜𝑙 𝑝𝑣 = 𝐴𝑐𝑠 𝐿 𝑐𝑜𝑟𝑟 𝑒𝑐𝑡𝑒𝑑 = 𝐴𝑐𝑠 (𝐿 − 2𝑅) (1)
𝐴𝑐𝑠 = 𝑊 𝐻 − 4𝑅 2 + 𝜋𝑅 2 − (𝑊𝑖𝑛𝑛𝑒𝑟 𝐻𝑖𝑛𝑛𝑒𝑟 − 4𝑅𝑖𝑛𝑛𝑒𝑟
2 2
+ 𝜋𝑅𝑖𝑛𝑛𝑒𝑟 ) (2)
Where 𝑊 is the width of the vessel, 𝐻 is the height, 𝐿 is the length, 𝑅 is the radius, 𝑊𝑖𝑛𝑛𝑒𝑟 is the width minus twice
the thickness, 𝐻𝑖𝑛𝑛𝑒𝑟 is the height minus two times the thickness, and 𝑅𝑖𝑛𝑛𝑒𝑟 is the radius minus the thickness. The
caps’ volume calculation similarly accounts for the different layers of the vessel.
The mass of the pressure vessel (𝑚 𝑝𝑣 ) is determined using Eq. (3) where the products of each layers volume (𝑉𝑖 )
with it’s corresponding material density (𝜌𝑖 ) are summed together.
𝑛
𝑚 𝑝𝑣 = Σ𝑖=1 𝑉𝑖 𝜌𝑖 (3)

There are two constraints placed on the pressure vessel system. If either is violated, then the corresponding set of
inputs is marked as a failed case. A minimum thickness constraint was implemented to ensure the safety of the pressure
hull. The minimum thickness (𝑡 𝑚𝑖𝑛 ) was calculated using Eq. (4), where 𝑃 is the internal pressure (given by ECLSS), 𝑅
is the radius, 𝑆 is the yield stress of the chosen material, and 𝐸 is the joint efficiency (set to 1 because the vessel is
assumed defeatured). If the input configuration consisted of a set of layers that were too small, then the constraint was
violated.
𝑃𝑅
𝑡 𝑚𝑖𝑛 = [11] (4)
2𝑆𝐸 + 0.4𝑃
The second constraint came from the ECLSS model. The ECLSS model calculates a minimum required internal
volume from the mission’s necessary equipment and how much volume the crew requires. If the given pressure vessel
geometry results in a volume less than this minimum, the constraint is violated. This minimum volume constraint varies
heavily with the system inputs and thus was different for each case.

3
 B. Chassis
The chassis of a vehicle is responsible for providing structural stability by holding all the subsystems and components
together. To maintain the stability, it must withstand the majority of the dynamic weight loads without any failures.
Depending on the intended use of the vehicle, the design of the chassis may vary. The model outlined here supports two
of the primary chassis types, ladder and backbone [16]. Examples of what these frames can be found in Fig. 4-7.

13 14
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Last Updated: Aug 05, 2022 Last Updated: Aug 05, 2022
ASDL Intellectual Property ASDL Intellectual Property

Fig. 4 Backbone chassis with four wheels. Fig. 5 Backbone chassis with six wheels.

17 16

Last Updated: Aug 05, 2022 Last Updated: Aug 05, 2022
ASDL Intellectual Property ASDL Intellectual Property

Fig. 6 Ladder chassis with four wheels. Fig. 7 Ladder chassis with six wheels.

The chassis interacts with every component of the rover and experiences the gamut of stress states during a mission.
As such, a usable parametric model has to make certain assumptions in order to produce relevant results. Taking this all
into consideration, the model focuses on producing a realistic mass estimate for the chassis rather than conducting, the
more popular, extensive Finite Element Analysis (FEA) based analysis [17–20].
In service of this, some key chassis assumptions are baked into the model. First, the chassis shape was defeatured
and discretized into segments (see Fig. 8), and all the segments had the same cross-section: a hollow rectangle with
8mm wall thickness. Second, the density of the material was the same for all the segments. Third, the effects of heat
transfer and radiation were not considered. Finally, a correction factor on the mass calculation would account for the
mass of the connections and features.
The overall structure of the chassis was determined by longitudinal and lateral beams, and similar types of beams
were interconnected by connection beams, depending on the number of required beams. The lateral beams could adapt
their shapes to form saddles for different shapes of pressure vessels (see Fig. 8-9).

4
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Fig. 8 CAD model of the chassis for four wheels. Fig. 9 CAD model of the chassis for six wheels.

First the volume of the chassis is found by adding the volume of each beam. The volume of each beam (𝑉 𝑜𝑙 𝑖 )
was found by multiplying it’s cross-sectional area (𝑎 𝑐 ) by the length of the beam (𝐿 𝑖 ). For curved beams, a numerical
approach was used to integrate Eq. (5), the derivative of the curve (𝑑𝑦/𝑑𝑥) was found using the finite difference method.
Equation (5) was determined by the creation of a modular function that took values of the pressure vessel geometry and
converted it into an equation for a circle.
∫ 𝑥 𝑓 √︃
𝐿𝑖 = 1 + (𝑑𝑦/𝑑𝑥) 2 𝑑𝑥 (5)
𝑥0

The mass, as shown in Eq. (6), was then calculated by summing across each beams volume multiplied with the
density (𝜌). Note that 𝑝 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑜𝑟 𝑠 is a ratio representing the mass of a connector over the entire weight of the frame
without connectors. The factor is design specific and hence varies from case to case. For this case, the correction factor
was determined with the assistance of a CAD model. 𝑁 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑜𝑟 𝑠 is the number of connectors in the frame.
𝑛
𝑚 𝑐ℎ𝑎𝑠𝑠𝑖𝑠 = (Σ𝑖=1 𝑉 𝑜𝑙 𝑖 𝜌𝑖 ) (1 + 𝑝 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑜𝑟 𝑠 ∗ 𝑁 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑜𝑟 𝑠 ) (6)

C. Suspension System
The suspension system is a critical component of the rover’s motion and one of the most complex systems to model.
Since only one rover has been deployed on the moon [21], and it was not pressurized, and many other rover designs only
exist on paper, there is a lack of historical data on lunar rover suspension masses. Further, most data on chassis design
and specifications is proprietary and any databases would take exorbitant effort. Therefore, this paper outlines a hybrid
mass estimator for the suspension system. The model breaks down suspension to its primary components, sizes them
separately, and then sums them together to get a total mass estimate. The estimator is summarized in Fig. 10.
The suspension model utilized several assumptions to ensure the estimator was parametric. First the cross-section
area of the control arms is a constant filled rectangle. Second, the model only considers static loads, and no dynamic
analysis was conducted. Third, dust mitigation mechanisms, radiation, and temperature effects are not considered
in the model. Fourth the model could only size McPherson Strut, Double Wishbone, and Trailing Arm suspension
configurations.
The inputs for the model were the sprung mass (obtained from the other systems), the vehicle and chassis dimensions,
the number of wheels, the wheel deflection, the installation ratio, the angle for the spring, the dimensions for the control
arms, and the material properties for aluminum and steel. The modeling process began by computing the volume
and mass of the control arms. This is done by multiplying the length of the arm with the cross section area, and then
multiplying the result (the arm volume) with the density of the indicated material to get the total mass.
Next, the mass of the shock absorber is iterated using an initial guess for the total mass of the shock and spring
(𝑀𝑡𝑜𝑡 𝑎𝑙 𝑔𝑢𝑒𝑠𝑠 ). During the shock-mass iteration, the Tresca failure criterion is employed to determine the appropriate
size of the coil spring. Once a guess is made, arbitrarily, the mass of the spring has to be estimated.

5
 21

Analytical Control Arms Model:

• Computes the volume and mass

Analytical Spring Model:

• Updates spring mass based on rover's total mass
• Checks the failure criterion

Regression Shock Model:

• Determines shock’s mass based on the ID of the spring
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Bending Stress Analysis on the Control Arms:

• Computes max stress
• Checks the failure criterion

Masses

Fig. 10 Diagram depicting the flow of the suspension system model.

In the spring loop, the mass exerting force on the front wheels is calculated using Eq. (7) [22] where 𝑚 𝑠 𝑝𝑟𝑢𝑛𝑔 is the
sprung mass, 𝑏 is the distance from the center of mass to the back wheels, and 𝑎 is the distance from the center of mass
to the front wheels. Then, the force felt by the spring (𝐹𝑠 ) is computed by considering the gravity (𝑔), installation ratio
(𝐼 𝑅), and installation angle for the spring (𝜃), see Eq. (8).
𝑚 𝑠 𝑝𝑟𝑢𝑛𝑔 ∗ 𝑏
𝑚1 = (7)
𝑏+𝑎
𝑚 𝑡𝑜𝑡 𝑎𝑙 𝑔𝑢𝑒𝑠𝑠𝑒𝑑 ∗ 𝑔
𝐹𝑠 = (8)
𝐼 𝑅 ∗ 𝐶𝑜𝑠(𝜃)
The next step is to determine different suspension metrics such as the spring rate (𝐾𝑠 ), wheel rate (𝐾 𝑤 ), ride rate
(𝐾𝑟 ), natural frequency (𝜔), and Wahl’s factor (𝐾). The Wahl’s factor takes into account the effect of direct shear and
change in coil curvature [23]. Equations (9)-(13) show the computations, where 𝛿 is the spring maximum deflection, 𝐶
is the spring index, and 𝑘 𝑡 is the tire rate.
4 ∗ 𝐶 − 1 0.615
𝐾= + (9)
4∗𝐶 −4 𝐶
𝐹𝑠
𝐾𝑠 = (10)
𝛿
𝐾𝑠
𝐾𝑤 = (11)
𝐼 𝑅 2 ∗ 𝐶𝑜𝑠(𝜃)
𝐾𝑊 ∗ 𝑘 𝑡
𝐾𝑟 = (12)
𝐾𝑊 + 𝑘 𝑡
√︂
1 𝐾𝑟
𝜔= (13)
2 ∗ 𝜋 𝑚1
The geometry of the spring is then calculated using Eqs. (14)-(18), where 𝐷 is the average diameter of the spring, 𝑑
is the diameter of the wire, 𝜏𝑚𝑎𝑥 is the maximum shear stress (this is where the Tresca failure criterion was implemented

6
 by setting 𝜏𝑚𝑎𝑥 to have the ultimate stress), 𝑁 is the number of active coils, 𝑁𝑡 is the number of total coils, 𝐺 is the
shear modulus, and 𝜌 is the density of the spring material.
𝐷 𝜏𝑚𝑎𝑥 𝜋
= (14)
𝑑3 8𝐹𝑠 𝐾
𝐷
𝐶= (15)
𝑑
𝛿𝐺𝑑 4
𝑁= (16)
8𝐹𝑠 𝐷 3
(𝜋𝑑) 2 𝐷𝑁𝑡
𝑉𝑠 𝑝𝑟𝑖𝑛𝑔 = (17)
4
𝑚 𝑠 𝑝𝑟𝑖𝑛𝑔 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 = 𝜌𝑉𝑠 𝑝𝑟𝑖𝑛𝑔 (18)
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Next the shock absorber mass is computed using the calculated spring inner diameter. The spring mass is input into
one of four regression equations fit to sample shock data from similarly sized vehicles [24, 25]. Equation (19) is used for
diameters smaller than 0.075 m, Eq. (20) for diameters between 0.075 m and 0.095 m, Eq. (21) for diameters between
0.095 m and 0.13 m, and Eq. (22) for diameters greater than 0.13. The shock absorber mass is then fed back into the
total guess and the loop begins again. Eventually a converged value for the mass of the spring-shock system is found.

𝑚𝑎𝑠𝑠 = 30.645 ∗ 𝐷 + 0.297 (19)

𝑚𝑎𝑠𝑠 = 60.517 ∗ 𝐷 − 2.7002 (20)

𝑚𝑎𝑠𝑠 = 30.645 ∗ 𝐷 + 0.297 (21)
𝑚𝑎𝑠𝑠 = 34.487 ∗ 𝐷 − 0.267 (22)
Finally a static bending stress analysis is conducted on the control arms. If the calculated stress is below the
maximum allowable stress, the suspension mass is finalized and the model reports it’s results. If not, the model repeats
or throws an "unable to converge" error, see Fig. 10 for reference.

III. Model Validation

Two conceptual cases were used to validate the models proposed in this paper. The first case used a lunar rover
model originally designed by NASA and the Virginia Polytechnic Institute in 1992 [2]. The second case used the
Electric Lunar Rover (LER) developed by NASA for lunar exploration [8].
The pressure vessel model was compared first to hand calculations and a CAD model, with discrepancies between
this paper’s model and the CAD model being less than 1%. The overall dimensions of the rover, from [2], were a length
of 11 m, a radius of 4 m, a circular cross section, and 5 insulation layers. The model outlined in Section II.A calculated
a vessel mass of 1,613 kg, a 3% difference to the published mass of 1,564 kg, see Table 1. Similarly, the chassis was
closely modeled to the specification in [2]. The model presented a mass of 551 kg, in contrast to the 652 kg stated in the
referenced paper. This difference of 15% is not a cause for concern due to the high level that these models are intended
to be used at. The model proposed in Section II.C resulted in a suspension system of about 59 kg per wheel, whereas
[2] reported a weight of 60 kg per wheel assembly. Major accommodations were needed to tune the model using the
correction factor, but the initial comparison demonstrated that the model could approximate the values of published
conceptual design given it’s specifications. Overall, the results shown in Table 1 indicate nothing of extreme concern
about the capability of the models shown in Section II.

Table 1 Comparison between the proposed models and the published masses of the conceptual lunar rover.

Proposed Model (kg) NASA + VPI [2] (kg) Difference (%)

Pressure hull 1,613 1,564 3.13
Chassis 551 652 15.5
Suspension 59 60 1.67

7
 For the second validation case, the LER was modeled with 4.5 meters in length, 3 meters in height, and a geometry
as close as possible based on [8]. The propulsion system was assumed fully electric, with an adequate array of batteries
for the mission duration. Table 2 presents some of its key design parameters, contrasting the actual total mass and the
resulted total mass of the full parametric tool discussed at the beginning of Section II.

Table 2 Design parameters of the LER compared against the results from the propsed models

Design parameter LER ASTRAEUS

Wheel diameter 0.640 m 0.610 m
Mission length 14 days 14 days
Chassis material Aluminum Aluminum
ECLSS Not included Not included
Suspension Independent Double wishbone
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Dimensions 4.5 m x 3 m 4.5 m x 3 m

Mass 3000 kg 3578 kg

The outlined models calculated a total mass of 3,578 kg, resulting in 19.3% error. This error was deemed acceptable
due to the unavailability of precise LER specifications and limitations on the details of the structural components.
Further, an ability to estimate final mass to within 20% of the actual final value before any designing even begins
would prove invaluable to relevant designers. The results of the two simulations indicate that the proposed models are
reasonably consistent with the models used for validation. This indicates that, at the incredibly early design stage it is
intended for, the tool can provide valuable insight into how a final design can pan out.

IV. Problem Setup

After the tool had been validated, an example requirements exploration was conducted to demonstrate the developed
capability. This involved varying a set of variables across multiple cases and then using the compiled results to analyze
varying requirements’ effect on the final design space.
The Design of Experiments (DOE) that was used for the analysis consisted of 10,000 cases with 25 input variables.
Table 3 shows the variables and the ranges they were tested across. The ranges themselves were set for a variety
of reasons, including historical data, context from other studies, and hypothesised goals for the Artemis Missions
[2–4, 10, 14, 20]. The converged value for all the cases can then be analyzed together to create surrogate models of the
relationships between the input and output variables. These surrogate models can then be used to explore how varying
requirements on the output variables effect the rover’s feasible design space. Note that some of the variables were varied
in discrete steps and some were considered continuous, each variable’s distinction is marked in Table 3.

A. Brief description of relevant models

Table 3 has some ranges and variables that need explanation, even if they are not the core of the analysis done in
Section V. The variables most worth describing relate directly to four models referenced in Fig. 1, Ingress/Egress,
Science, Navigation, and Communications and Data Handling (C&DH). Each of these models are described briefly here
to help describe what the sample analysis is about.

Ingress/Egress There were four mutually exclusive options for Ingress and Egress. these were a suit-lock (sometimes
referred to as a suit-port), an air-lock, an integrated combination of the two called a suit-air-lock, and a combination of
the two that doesn’t integrate them together. This means that there are three different items that need to be sized. The
sizing for the air-lock was based off of values found in [26]. The suit-lock was sized using [27] to scale the air-lock
values from [26]. This scaling was done to account for the fact that the air-lock now has to get larger as the crew
increases because each suit needs to fit on to the wall. Finally the integrated and non-integrated combined choices
were made through inference and further analysis of the data in [27] and [26]. Relative to Table 3, the options were
changed by discreetly choosing their corresponding number. The Air-Lock only option was number 1, Suit-Lock only
was number 2, the integrated Suit-Air-Lock was assigned number 3, and the non-integrated combination was number 4.
As shown in the table, the DOE only explored the first three options as the fourth would always be the heaviest and
hence was not irrelevant to the preliminary analysis.

8
 Table 3 Variables and corresponding ranges for the DOE used in the sample analysis

Variable Description Min Max Discrete?

Mission Length (hr) Time spent without resupply 24 672
Crew Size Number of Crew 2 8 Y
Fuel Cell-Battery Ratio Power supplied 𝐵𝑦𝐹𝑢𝑒𝑙𝐶𝑒𝑙𝑙𝑠
𝐵𝑦𝐵𝑎𝑡𝑡𝑒𝑟𝑖𝑒𝑠 0.0 1.0
Acceleration (m/𝑠2 ) Rover peak acceleration 0.5 3.0
Number of Wheels - 4 10 Y
Corner Radius Ratio How pill-shaped the PV is 0.1 0.5
Height (m) Height of rover 2.2 5.0
Width (m) Width of rover 2.0 5.0
Length (m) Length of rover 3.0 10.0
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

CDH Choice Comms. & Data Handling method 1 2 Y

Ingress/Egress Choice Method for Ingress/Egress 1 3 Y
ECLSS Backups # of ECLSS system backups 0 1 Y
CO2 Removal Option CO2 removal method 0 1 Y
O2 Generation Option Generate Oxygen (0) or bring a tank (1) 0 1 Y
H2 O Management Option Include water recovery (0) or not (1) 0 1 Y
Chassis Beam Height (m) Height of chassis beam cross section 0.01 0.10
Chassis Beam Width (m) Width of chassis beam cross section 0.01 0.10
Number of Chassis Saddles Number of curved chassis beams 1 5 Y
Number of Saddle Connectors Number of connectors between saddles 1 7 Y
Wheel Radius (m) Radius of rover tire 0.25 .75
Suspension Installation Ratio Shock placement along control arm 0.50 1.00
Wheel Deflection (m) Maximum wheel deflection 0.05 0.25
Control Arm Length (m) Length of the control arms 0.25 1.00
Control Arm Height (m) Height of control arm cross section 0.05 0.15
Control Arm Width (m) Width of control arm cross section 0.05 0.15

Science The science model for this version of the tool was a simple rubberized payload of a robotic arm. The values
of this arm were based off of the technical specifications found in [28] with some margins applied to account for the fact
that it’s on the moon and needs to be integrated. The final values of the payload were a mass of 120 kilograms, power
draw of 300 Watts, and nominal voltage of 120 volts. These values can be changed at will for different missions if
necessary.

Communication and Navigation For this stage of the tool, communication and navigation systems were less important,
and only appeared as a singular choice in the tool (CDH Choice in Table. 3). This is because the size, mass, and
power usage of the systems do not vary with mission changes. If the mission is twice as long, the communication and
navigation systems stay the same. The two communication systems modeled in the tool were a phased array antenna and
a dish antenna modeled after the Lunar Roving Vehicle’s LCRU [21]. The phased array antenna being used in this
tool is a tile technology based array with a 45x15 tiled design [29]. Relative to Table 3, the phased array was assigned
number 1, and the LCRU assigned number 2.
For navigation, the model used comes from the technology used on the lunar roving vehicle from the 1960’s. This
system has a mass of 5 kg and a power required of 60 watts [21]. The total mass, volume, and power required for all
these models were all used by the openMDAO framework to then provide load cases and size requirements for the power
and structural components.

9
 B. Constraints for the Analysis
Once the results of the 10,000 cases are compiled and the surrogate models are created, the requirements analysis
needs some constraints to be conducted. These are minimum or maximum values for certain outputs that limit the
design space. The constraints chosen, as well as where their values were sourced, for this paper’s analysis are shown in
Table 4. The constraints are loosely based around a hypothetical requirements study to determine if the Apollo Lunar
Module could be replaced with a pressurized lunar rover. As such, the total mass and total volume constraints are
approximates of the lunar module and Saturn V’s specifications. The values themselves are arbitrary and the use case is
just a thought experiment, but an experiment that can showcase the developed capability.

Table 4 Constraints used for the requirements analysis sample problem

Constraint Value Reason
Total Mass 12000 kg ≈ mass of Apollo Lunar Module without ascent stage [30]
Total Volume 170 m3 ≈ 2x volume of Apollo Lunar Module [30]
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Sprung Mass 11300 kg ≈ Total Mass ∗ 95%

Suspension Mass 115 kg ≈ Total Mass - Sprung Mass
# of Wheels
Chassis Mass 300 kg Visibility on Contour Plot
Pressure Vessel Mass 1470 kg Visibility on Contour Plot
Pressure Vessel Vol 95 m3 Visibility on Contour Plot

V. Results and Discussion

Total Mass – 12,000 kg

Total Volume – 170 m3

Sprung Mass – 11,300 kg

Suspension Mass – 115 kg

Chassis Mass – 300 kg

Pressure Vessel
Mass – 1470 kg

Pressure Vessel
Volume – 95 m3

Fig. 11 Contour Plot showing the rover’s design space based on the analysis conducted. The constraints are set
arbitrarily based on the Saturn V’s fairing size [31] and visibility on the plot.

The analysis’ results are displayed using a contour plot, shown in Fig. 11. This contour plot uses two of the most
important mission requirements as the x and y axes, Crew Size and Mission Length respectively. Both influence nearly
every subsystem in a rover from the required motor torque to how many space suits there should be. These variables are
arguably the first requirements that decision makers would want to determine when approaching a rover design. The
results are analyzed using a least squares based fitting method to map the inputs to the outputs. Each line has a feasible
(blank) side and a violates the constraint (shaded) side. By overlaying all the constraints together, a region of "feasible"
rovers is found that pass all the constraints. Designers can then use that region to determine overall rover requirements.
For this paper, the contour plot shows 7 limits that are directly based on the structural model. The analysis conducted is
not definitive and is just to showcase the developed capability.

10
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

Fig. 12 Contour Plot showing how the Total Mass constraint line shifts with ECLSS Backups and the constraint
value.

Each line, however, will change based on constant values set for the remaining 23 input variables, even if the
constraint value is constant. This results in interesting relationships between the constraints and the feasible space. For
example, the total mass line in Fig. 11 shows a limit of 12,000 kg. Figure 12 shows that if that limit is changed to 13,000
kg then the line will move up and to the right, while setting it to 11,000 kg will move it down and left. Further, if the
number of ECLSS backups is reduced from 1 to 0, the line will move up and right; so much so that 12,000 kg with 1
backup is less constraining than 11,000 kg with 0 backups. A result that requires explanation but is logically sound. A
rover that can’t exceed 12,000 kg but needs to have 1 ECLSS backup will have less "space" for crew and power systems.
While having a limit of 11,000 kg but not requiring any ECLSS backup means more weight can be taken up by crew
and power systems. Hence, a 12,000 kg max weight with no ECLSS backup is less constraining than 12,000 kg with
1 backup. This behavior is true for all the metrics to some capacity. To account for this Table 5 shows what input
constants were set for the displayed results.

Table 5 Constants used for generating the displayed contour plots.

Variable Constant Variable Constant

Height (m) 4 Control Arm Length (m) 0.5
Width (m) 4 Control Arm Height (m) 0.1
Length (m) 7 Control Arm Width (m) 0.1
FuelCell-Battery Ratio 1 (only fuel cells) Wheel Radius (m) 0.5
Corner Radius Ratio 0.2 Acceleration (m/𝑠2 ) 1
Number of Wheels 6 Suspension Installation Ratio 0.75
CDH Choice 1 (Tile Based Phased Array [29]) Wheel Deflection 0.125
Ingress/Egress Choice 3 (SuitLock [26, 27]) Number of Chassis Saddles 3
ECLSS Backups 1 Number of Saddle Connectors 4
CO2 Removal Option 1 (4-Bed Molecular Sieve [32]) Chassis Beam Height (m) 0.05
O2 Generation Option 0 (Tank Storage) Chassis Beam Width (m) 0.05
H2 O Management Option 1 (No Recovery)

11
 There are a few key constraints that can be seen following the same trends as the Total Mass. These are the Total
Volume, Sprung Mass (the weight of all components supported by the suspension), and Suspension Mass. All of these
parameters, as shown in Fig. 11, indicate that increases in either Crew Size or Mission Length will take the design into
their shaded regions. This means that as the Rover’s crew or mission length increases it’s mass, volume, sprung mass
and suspension mass will all increase. This relationship is incredibly intuitive and can easily be compared against one
another to help designers understand what the most constraining factor is. For example, it’s possible a rocket’s fairing
volume causes more issue than its lifting capacity.
Putting the constraints all together, the contour plot defines a feasible space for the rover design based on the set
constants and design parameters. Looking at Fig. 11, it can be seen that the maximum possible Mission Length is
approximately 625 hours with a Crew Size of 5. While the maximum Crew Size of 6 can only have a Mission Length of
100 hours. Any combination below and to the right of these is considered feasible based on this analysis. Going back to
the thought experiment from Section IV, the feasible space constraints what kind of rovers can be used to replace the
Apollo Lunar Module. According to these results, a rover capable of carrying a crew of 4 for 600 hours without resupply
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

can be taken up to the moon using the Saturn V by simply replacing the slot where the Apollo Module originally sat.
These results’ asterisk, summarized in Tables 4 & 5, has to be considered when making conclusions. Because of
this asterisk, it is unlikely a 6 Crew rover can actually take the lunar module’s slot in the Saturn V and make it to the
moon. The results do, however, wonderfully exemplify the the capability developed in this paper. Depending on the
constraints put into the tool, engineers can easily use results like these to set very solid and physics based requirements
for their designs.

VI. Conclusion
In summary, the implemented Python models showed promising results for sizing pressurized lunar rovers. The
pressure vessel model exhibited high accuracy when compared to hand calculations and a CAD model, with discrepancies
of less than 1%. The different components of the structural model provided reasonable estimates when compared to
published masses of conceptual designs, with percent differences of about 4%, 15%, and 2%, for the pressure vessel,
chassis, and suspension system. Despite some challenges encountered in the integration of all subsystems, the models
demonstrated the capability to obtain inputs from a JSON file and share values across subsystems. Further investigations
are still ongoing to address an unexpected circular region where no feasible design was possible. Additionally, there are
ongoing efforts to add stress analysis capabilities to the chassis model. Overall, the developed structural parametric tool
provides a valuable contribution to the design and optimization of pressurized lunar rovers, facilitating the exploration
of the Moon and paving the way for future space exploration missions.
There are numerous areas of potential exploration after the creation of this tool. As indicated in the results, there are
issues with the pressure vessel and the chassis mass that can be the subject of future exploration. Further development
of the non-structural models would also greatly improve the fidelity of the tool. More extensive analysis can conducted
of specific areas of the design space. The tool can even be integrated with other models to conduct more intricate and
extensive lunar infrastructure studies. Overall, the creation of this tool opens many avenues for more in-depth analysis.

Acknowledgments
The authors would like to thank Claire Harding, Thomas Leroy, and Varick Peak for their contributions to the
modeling and creation of the tool, as well as for being excellent team members. This work was supported by Jacobs
Technology and NASA Johnson Space Center.

References
[1] NASA, “NASA’s Lunar Exploration Program Overview,” Tech. rep., NASA, 2020. URL https://www.nasa.gov/wp-
content/uploads/2020/12/artemis_plan-20200921.pdf.

[2] Bhardwaj, M., Bulsara, V., Kokan, D., Shariff, S., Svarverud, E., and Wirz, R., “Design of a Pressurized Lunar
Rover,” Tech. rep., Department of Aerospace and Ocean Engineering, Virginia Polytechnic Institute and State University,
04 1992. URL https://www.researchgate.net/profile/Richard-Wirz/publication/24302438_Design_of_a_
pressurized_lunar_rover/links/54cff2830cf29ca81100aede/Design-of-a-pressurized-lunar-rover.pdf.

[3] Collins, C. W., Gomez, A., Muñiz, R., Musson, D. M., and Fowler, W. T., “Conceptual Design of A Mars Surface Transportation
System (MSTS),” Second Annual HEDS-UP Forum, 1999, p. 201.

12
 [4] Damer, B., Rasmussen, D., and Newman, P., “Design simulation in support of NASA’s robotic and Human Lunar Exploration
Program,” Space 2006, 2006. https://doi.org/10.2514/6.2006-7397.

[5] Chun, W., Kirkland, J., Conrad, K., Brady, D., Fisher, R., and Fernandez, K., “Lunar Surface Mobility,” Space, 2006.
https://doi.org/http://doi.org/10.2514/6.2006-7396.

[6] Cohen, M. M., “Mobile Lunar Base Concepts,” AIP Conference Proceedings, Vol. 699, No. 1, 2004, pp. 845–853.
https://doi.org/10.1063/1.1649649, URL https://doi.org/10.1063/1.1649649.

[7] Toyota, “Small Pressurized Rover Prototype,” Tech. rep., Toyota Motors Corporation, 2022. URL https://global.toyota/pages/
global_toyota/mobility/technology/lunarcruiser/lunarcruiser_01_en.pdf.

[8] Litaker, H., Thompson, S., and Howard, R., “Human Habitation in a Lunar Electric Rover during a 14-Day Field
Trial,” Vol. 54, No. 19, 2010, pp. 1459–1463. https://doi.org/10.1177/154193121005401922, URL https://doi.org/10.
1177/154193121005401922, publisher: SAGE Publications Inc.
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

[9] Abercromby, A., Gernhardt, M., and Litaker, H., Desert Research and Technology Studies (DRATS) 2008 eval-
uation of Small Pressurized Rover and unpressurized Rover prototype vehicles in a Lunar analog environment,
2010. URL https://www.researchgate.net/publication/259755761_Desert_Research_and_Technology_Studies_DRATS_2008_
evaluation_of_Small_Pressurized_Rover_and_unpressurized_Rover_prototype_vehicles_in_a_Lunar_analog_environment.

[10] Milliken, W. F., Milliken, D. L., and Olley, M., Chassis Design: Principles and Analysis, Society of Automotive Engineers,
2002. URL https://api.semanticscholar.org/CorpusID:107668946.

[11] Moss, D. R., Pressure Vessel Design Manual, Gulf Professional Publishing, 200 Wheeler Road, Burlington, MA 01803, USA,
2004.

[12] Milliken, W. F., and Milliken, D. L., Race Car Vehicle Dynamics, Society of Automotive Engineers International, 1994.

[13] Harding, C., and Peak, V., “Power and Thermal System Modeling for a Pressurized Lunar Rover,” AIAA SciTech 2024, 2024.

[14] Center, N. J. S., “Advanced Life Support Sizing Analysis Tool (ALSSAT) version 12.0(MSC-25510-1),” NASA Technology
Transfer Program, 2023. URL https://software.nasa.gov/software/MSC-25510-1.

[15] Gray, J., Moore, K., and Naylor, B., OpenMDAO: An Open Source Framework for Multidisciplinary Analysis and Optimization,
2012. https://doi.org/10.2514/6.2010-9101.

[16] Ali, R. D., “Design and Development of a Chassis Concept for an Autonomous Airport Shuttle,” Tech. rep., KTH Royal Institute
of Technology School of Engineering Science, Stockholm, Sweden, 2020. URL https://www.diva-portal.org/smash/get/diva2:
1465541/FULLTEXT01.pdf.

[17] Kiran, L., Kakkeri, S., and Deshpande, S., “Proposal of Hybrid Composite Material for Light Commercial Vehicle Chassis,”
Materials Today: Proceedings, Vol. 5, No. 11, Part 3, 2018, pp. 24258–24267. https://doi.org/https://doi.org/10.1016/j.matpr.
2018.10.221, URL https://www.sciencedirect.com/science/article/pii/S2214785318325161.

[18] Nandhakumar, S., Seenivasan, S., Mohammed Saalih, A., and Saifudheen, M., “Weight optimization and structural analysis
of an electric bus chassis frame,” Materials Today: Proceedings, Vol. 37, 2021, pp. 1824–1827. https://doi.org/https://doi.
org/10.1016/j.matpr.2020.07.404, URL https://www.sciencedirect.com/science/article/pii/S2214785320355140, international
Conference on Newer Trends and Innovation in Mechanical Engineering: Materials Science.

[19] Muzakkir Ahamed, M., and Natrayan, L., “Dynamic analysis and structural simulation of aluminum composite material based
automobile chassis,” Materials Today: Proceedings, Vol. 62, 2022, pp. 2244–2249. https://doi.org/https://doi.org/10.1016/j.
matpr.2022.03.466, URL https://www.sciencedirect.com/science/article/pii/S2214785322018156, international Conference on
Design, Manufacturing and Materials Engineering.

[20] Pitre, R. R., “Systematic Structural Optimization of a Next Generation Lunar Rover Chassis,” Ph.D. thesis, Queen’s University,
Ontario, Canada, 2015. URL https://bac-lac.on.worldcat.org/oclc/1033003888.

[21] Lund, T., Early Exploration of the Moon: Ranger to Apollo, Luna to Lunniy Korabl, Springer International Publishing, Cham,
2018. https://doi.org/10.1007/978-3-030-02071-2.

[22] Vatsal, J., “Coilover Shock absorber | Coil spring designing | Damper calculations | Mounting points of coilover,” https:
//www.oxfordreference.com/display/10.1093/acref/9780198832102.001.0001/acref-9780198832102-e-7256, Retrieved 28
November 2023.

13
 [23] Walh, A. M., Mechanical Springs, McGraw-Hill, (2), 1963.

[24] FordParts.com, “Suspensions,” https://parts.ford.com/shop/en/us/suspension/ride-control/shock-absorber-assy-9763835-1?

pdp=y, Retrieved 28 November 2023.

[25] Motors, G., “ACDelco:shocks,” https://parts.gmparts.com/search?facet=ads_f12501_ntk_cs:%22OE%22&searchTerm=Shocks,

Retrieved 28 November 2023.

[26] Griffin, B. N., “Lunar Habitat Airlock/Suitlock | Earth ; Space 2008 - ASCE library,” ASCE Library, 2012. URL https:
//ascelibrary.org/doi/abs/10.1061/40988%28323%29102.

[27] Fullerton, R. K., “Eva Tools and Equipment Reference Book - NASA Technical Reports Server (NTRS),” Tech. rep.,
NASA-TM-109350, Nov 1993. URL https://ntrs.nasa.gov/citations/19940017339.

[28] UNIVERSAL ROBOTS, “UR20 IP IP54,” Tech. rep., Universal Robots, 2023. URL https://www.universal-robots.com/media/
1824603/ur20_data_sheet.pdf.
Downloaded by Georgia Institute of Technology on October 24, 2024 | http://arc.aiaa.org | DOI: 10.2514/6.2024-0768

[29] Herd, J. S., and Fenn, A. J., “Design considerations for space-based radar phased arrays,” IEEE MTT-S International Microwave
Symposium Digest, 2005, 2005, pp. 1631–1634. https://doi.org/10.1109/MWSYM.2005.1517019.

[30] Roberson, F. I., Hanley, J. B., and Eichelman, W. F., “Apollo 11 Lunar Module / EASEP,” Tech. Rep. 1969-059C, NASA Space
Science Data Coordinated Archive (NSSDCA), Washington, DC, 1969.

[31] Wade, M., “Saturn V,” http://www.astronautix.com/s/saturnv.html, 2019.

[32] Knox, J. C., Mulloth, L. M., and Affleck, D. L., “Integrated Testing of a 4-Bed Molecular Sieve and a Temperature-Swing
Adsorption Compressor for Closed-Loop Air Revitalization,” https://ntrs.nasa.gov/api/citations/20040085894/downloads/
20040085894.pdf, 2004.

14