Development and test of a Lunar Excavation and Size Separation System (LES3) for the LUVMI‐X rover platform

Future sustained human presence on the Moon will require us to make use of lunar resources. This in‐situ resource utilisation (ISRU) process will require suitable feedstock (i.e., lunar regolith) that has been both acquired and prepared (or beneficiated) to set standards. Acquisition of pre‐processed regolith, is an often overlooked engineering challenge in the demanding and low‐gravity environment of the lunar surface. Currently, regolith excavation and size separation are often developed independently of each other. Here, we present the Lunar Excavation and Size Separation System (LES3), which is an engineered one‐system solution to combine the acquisition of lunar regolith as well as separate it into two distinct size fractions, and therefore, can assist to define the quality of the feedstock material for ISRU processes. Intended for use with a lightweight (40–60 kg) lunar rover (LUnar Volatiles Mobile Instrumentation‐X; LUVMI‐X) currently under development, the mechanism utilises vibrations to reduce excavation forces and facilitate size separation. Low excavation forces are crucial for lunar excavators to be deployable on lightweight robotic platforms as limited traction forces are available. The rationale behind the mechanism is explained, its capabilities in the support of science and ISRU are showcased, and results from several laboratory test campaigns, including tests of gravitational dry sieving of different regolith simulants, are presented. The LES3 can excavate up to 100 g in a single charge while maintaining excavation forces of less than 8 N and having a mass of less than 2 kg. Finally, areas of improvement for a second iteration of the design are presented and explained. The LES3 proof of concept shows that combining of regolith excavation and size‐separation in a single mechanism is feasible.


| INTRODUCTION
The Moon and its exploration have once again become an important aspect of most space agency roadmaps, which in turn acts as motivation for a multitude of private companies to develop launch vehicles, landers, and exploration technology (Chavers et al., 2016;Reddy, 2018;von Ehrenfried, 2020;Voosen, 2018).
Most ISRU applications that are currently under investigation in a terrestrial laboratory setting use regolith analogues or simulants which have been sieved to a certain particle size distribution (PSD) before experiments are conducted (Taylor et al., 2016). While a process requiring a well-defined PSD or a small maximum particle size is relatively simple to achieve in a laboratory on Earth with different methods like wet and dry sieving, it becomes challenging in a lunar environment. In most publications detailing ISRU applications, this fact is not acknowledged and having access to a preprocessed feedstock is assumed a given. Table 1 shows an overview of certain ISRU applications and the level of reported feedstock preprocessing. Table 1 shows that most potential end-users of excavated regolith material rely on or benefit from a certain level of size separation. Currently, excavation and size separation are usually considered as two independent steps in the ISRU process chain (see Just et al., 2020b, for more details on the ISRU process chain), requiring regolith transport between the different processing sites or mechanisms. Thus, the development of a mechanism combining regolith excavation and beneficiation into one system seems highly beneficial. LUVMI-X is a small and lightweight four-wheeled lunar rover currently under development by Space Applications Services (Garcet et al., 2019;Losekamm et al., 2021), of which a rendering can be seen in Figure 1. The rover has a total mass of 40-60 kg, and a total payload capacity of 24 standard units (U) (1 U ≈ 10 × 10 × 10 cm; Gatsonis et al., 2016), split into 12 U per payload bay (front and back of the rover). Therefore, any payload mechanism must be storable within this envelope for launch. The ground clearance of~30 cm when roving can be lowered to~10 cm with the use of its suspension, allowing mechanisms to be deployed closer to the lunar surface, reducing the necessary reach, and thus structural mass. Due to its low mass, the rover can only provide a limited amount of effective traction force, requiring any excavation subsystem to operate with minimal excavation forces.
The development of a low-mass and low excavation force mechanism capable of excavating and size separating the lunar regolith is the objective of the presented study. When incorporated into LUVMI-X, the LES 3 will have three main functions in support of scientific as well as ISRU activities on the lunar surface: • Excavation and feedstock beneficiation: Two distinct size fractions of regolith feedstock can be delivered to different ISRU processes, which is beneficial to the process control and product quality of a multitude of ISRU applications, like for instance additive manufacturing, regolith sintering, or oxygen extraction (see Table 1).
• LIBS support: For compositional analysis of the lunar surface, LUVMI-X carries a laser-induced breakdown spectroscopy (LIBS) system on-board (VOlatiles Identification by Laser Ablation (VOILA)) (Garcet et al., 2019;Losekamm et al., 2021;Vogt et al., 2020Vogt et al., , 2021. Our proposed excavation method will be able to scrape/trench into the surface of the lunar regolith, revealing subsurface rock and soil samples available for LIBS investigations. Therefore, soils of interest or specific soil features, such as water content, can be investigated in more detail or at different depths (Lasue et al., 2012).
• Geotechnical properties of regolith: High-resolution images of the excavated trenches, as well as the two size fractions within their storage containers or excavated piles can provide information about the particle size distribution as well as physical properties of the soil, such as friction angles and cohesion (by measuring the angle of repose; for more detailed descriptions of these techniques, see Moore et al., 1999;Sullivan et al., 2011). Additionally, these strength parameters can also be calculated from the recorded excavation forces (Kobayashi et al., 2006).
Here, it is important to differentiate between a sampling/excavation mechanism intended to support analytical scientific research, like, for example, drill excavation as used in ESA's PROSPECT payload (Sefton-Nash et al., 2018), and a mechanism for which the sole purpose is to support ISRU activities, such as we propose for LES 3 . For the former, cross-contamination of different sampling locations is an important issue that must be addressed and mitigated. The presented mechanism on the contrary, is intended and designed to support ISRU applications such as additive manufacturing or oxygen extraction, where cross-contamination of samples is less important. While LES 3 is directly capable of performing the first presented function, it supports the remaining two functions indirectly by excavating soil and its subsequent imaging.

| Working principle
The working principle of LES 3 consists of four main steps, which can be seen illustrated in Figure 2 and which will be explained in more detail below: 1. Accumulation of soil: After the arm has been lowered, the rover pushes the inlet through the soil at a shallow angle (here 15 degree), as shallow angles have been proven to result in low excavation forces (Just et al., 2021). Both outlet ports are blocked by spring-loaded gates and the vibration motor is operating. After a sufficient time, indicated by the front area of the inlet being filled with regolith or an increasing surcharge mass, the vibration motor (details of vibration in Section 5) stops, and the mechanism moves into Position 2-the pre-separation phase.

Preseparation phase:
Since not all material is being pushed through the front sieving plate in Position 1, the second step is to pre-separate larger rock fragments (first step of size separation). In this position, which orientates the front sieving F I G U R E 1 Rendering of the LUVMI-X rover platform on the lunar surface. Image credit: Space Applications Services (publicly accessible at: https://www.h2020-luvmi-x.eu/gallery-page/)

| Mechanism design
The system consists of three distinct parts: The inlet or leading edge, the arm, and the base. The latter's design is expected to change, as the design of the turret is heavily depended on the way the mechanism will be integrated into the rover platform and the storage location for the excavated materials; specifications that are still being defined by the rover team at the moment. Thus, this part will only be discussed briefly in this publication. Figure  The excavation mechanism is based on a cylindrical inlet (see aperture size discussed in Section 5) separates out larger rock fragments and, therefore, acts as the first stage of regolith size separation. A finer mesh/sieving plate (see Figure 3 feature h; aperture size discussed in Section 5) will provide a second size separation step, leading to the creation of two distinct particle size fractions within the mechanism. Outlets for the coarse fraction (see Figure  The coarse size fraction will enter the arm through a chute (see attached to a base that acts as a turret with a third high-torque F I G U R E 2 Illustration of the working principle of the LES 3 mechanism in a lab environment (top) and as a schematic (bottom). The inlet is held in orange, the arm is blue, and black dots signal rotational axes. Rover as well as regolith planes are indicated. Soil is accumulated using the inlet system (1), pre-separated (2), the fine fraction is separated (3), and the coarse fraction is removed from the inlet (4). The illustration at the bottom shows and additional cleaning step (5). Videos of the mechanism in operation can be found at https://doi.org/10.48420/14511480, https://doi.org/10.48420/14510535, https://doi.org/10.48420/14511477, and https://doi.org/10.48420/14511483 JUST ET AL.
Therefore, the mechanism can be actuated in three active degrees of freedom in a yaw-pitch-pitch configuration, useful both for storing the mechanism during launch as well as to maximise its application possibilities. This assembly can be seen in Figures  More specifications of the mechanism, such as mass, power consumption, and excavated mass per scoop, can be found in Section 5, as they have been verified during the testing phase.

| DRY SIEVING OF REGOLITH
In principle, LES 3 utilises a two-stage vibrating sieve to achieve the necessary level of size separation. Dry gravitational separation of regolith is often considered challenging , and more complex separation techniques, such as electrostatic or magnetic separation, are proposed instead. While there are several publications on the separation of granular matter by means of vibration in a terrestrial setting (Kudrolli, 2004;Li & Tong, 2015;Wen et al., 2015), there is a very limited number of experimental studies available which investigate dry sieving of regolith or its simulants/ analogues (Wilkinson, 2011;Williams et al., 1979). It may be challenging to achieve separation down to very small particles in the challenging low-gravity lunar conditions and due to the cohesive nature (Mitchell & Houston, 1972) and electrostatic charging (Colwell et al., 2007)

| Dry sieving experimental set-up and methods
To inform a realistic level of size separation and, thus, define the aperture size for the proposed LES 3 excavation mechanism as well as to understand the time required to perform such separation, a standardised stand-alone dry sieving experiment was conceptualised.
The decision to perform this experiment as a stand-alone test is beneficial to the overall ISRU community, as the findings are not only applicable to the presented mechanism design but can inform decisions for a multitude of applications where regolith size separation is required. Figure 6 shows the experimental setup for this test. The Four different analogue materials were tested, to get an understanding of how particle shape and other material properties, for instance density or particle cohesion, affect the outcome and to allow for a more robust estimation of the feasibility of the intended ap- of the experiment, 235 or more measurements were flagged, the experiment was considered as "potentially over". If during the next 240 measurements 235 or more potential termination criteria were recorded, the experiment was ended, as no more recordable change in separated mass was to be observed. This method was applied to not underestimate the importance of small regolith particles in any ISRU application, where separation can take a considerable amount of time as the experiment gets closer to the separation limit. In other words, a long linear increase with a very shallow slope can add up to an appreciable mass. This becomes important for applications which require a particle size distribution larger than a certain cut-off, as too many fine particles in the feedstock can here reduce the quality of the ISRU product, cause an increased demand of consumables, or damage components, such as filters. After the experiment, both size fractions were weighed and the lost mass calculated. For all initial tests, sieves were cleaned with a sieve brush in between runs.

| Dry sieving experimental results
All performed experimental runs, including aperture size, used analogue, total analogue mass, mass of coarse/fine fraction, mass of lost material, percentage of passed material, expected percentage of passing based on the particle size distribution, as well as the required time to meet the end criterium (i.e., residence time) can be found in thus, using a sieve with an aperture coarser than 500 microns is untenable when considering the residence time. For the two data sets that are greyed out in Table 2, the residence time was interpolated based on their particle size distribution passing value due to the immense increase in time demand.  cessing mechanisms. It also shows that some of the particle size requirements for certain processes (Table 1) are difficult to achieve by dry sieving without manual sieve manipulation or large sieve shakers even in a terrestrial setting. Therefore, when developing new ISRU processes it is imperative to keep the necessary regolith pre-processing (i.e., sieving) requirements in mind. Additional testing performed with twice as much starting regolith simulant material (400 g) also showed clearly, that the quantity of material sieved at once (for an equivalent sieve area) should be kept to a minimum; an example of this can be seen in Figure 9. Once there is too much overhead (i.e., regolith simulant/analogue) in the sieve, which becomes consolidated due to the vibrations, the particles at the interface with the mesh cannot move sufficiently to orientate themselves in a way that would allow them to pass the sieve apertures. This results in passing percentages that are significantly lower than expected or drastically increased residence times (see Figure 9). Where the increase of screen area or decrease of batch size is not feasible, one solution to this problem could be the use of an additional soil agitation device, such as a rotating paddle or a wiper across the sieve surface, but this increases the complexity drastically (Singh, 2004).
This could, however, also help to clean the sieving plates after use. In the present experiments, vibrations with a large acceleration and low frequency were applied (see Section 4.1, for details), as this vibration mode was able to provide the necessary agitation of particles to allow for an efficient separation given the provided mass of analogue material. For more detailed discussions of the relation between to the platform (for set-up see Figure 6), rotated 90 degrees from the other motor. Both motors were operated at the same voltage and, therefore, frequency and acceleration. The test was performed with TUBS-M regolith simulant and the results can be seen in Table 3.   Table 3 shows that this reduces residence times significantly, as particles move horizontally as well as vertically across the screen, with the time savings becoming less prominent with a smaller aperture size. Thus, it is apparent that the vibration profiles of similar set-ups or mechanisms intended for the F I G U R E 6 Experimental setup for the dry gravitational sieving of different regolith analogue materials. The aim of this experiment was to investigate a feasible aperture size for use in the proposed mechanism. Individual components are: (a) extruded aluminum profiles, (b) 3D-printed platform, (c) standard 100 mm diameter woven-wire mesh test sieve, (d) 3D-printed clamps, (e) funnel (underneath sieve), (f) weighing container, (g) scale with beam loadcell, (h) vibration motor, (i) 3D-printed collar for vibration motor, (j) 3-axis accelerometer T A B L E 2 Overview of all performed dry sieving experiments including the used simulants, total analogue mass, aperture size of the used sieves, total duration of the experiment until the end condition was met, the coarse mass, the fine mass, the lost mass, the percentage passed, and the theoretical percentage passed based on the reported particle size distribution use on the lunar surface, need to be well characterised and subsequently optimised for each specific mission requirement; a task which complexity must not be underestimated.

| LES 3 LABORATORY TESTS AND PERFORMANCE CHARACTERISTICS
Based on the results explained in Section 4, for the application in  Note: The total analogue mass, aperture size of the sieves, total duration of the experiment until the end condition was met, the coarse mass, the fine mass, the lost mass, the percentage passed, and the time difference compared to the same experiment with one motor is listed.
F I G U R E 10 Acceleration profiles for dry sieving experiments with (a) one vibration motor and (b) two vibration motors. Dashed lines indicate the average (20) maximum/minimum acceleration of the most prominent vibrational axis (y-axis). Nominal acceleration of one motor (stand-alone) at 2.5 V was 4.5 g, as per datasheet. Axes directions can be seen in Figure 7 JUST ET AL.  Table 4  would have not made sense due to the operating principle of the mechanism, which includes a screen for size separation. Table 4 shows all experimental runs, the maximum horizontal excavation force F H_max , as well as the excavated mass. The distance traveled by the inlet through the sandbox was 31 cm for each run, resulting in an experiment time of around 31 s. Figure 11 (features a-c) shows the necessary horizontal excavation force profiles for the experiments detailed above.
We observe that the overall excavation forces are several times lower than the traction force provided by the rover (~20 N on the lunar surface; Just et al., 2021), crucial for the mechanism to operate successfully (Table 4 and Figure 11). The run without a front sieving plate displays the highest maximum forces overall across all runs, as here the material enters the inlet relatively unhindered (i.e., no screening out of any particle sizes), leading to a shorter surcharge length and advancement of material further into the inlet. Thus, the most material of all runs has to be displaced in this instance. Without a screen, there is less build-up of surcharge in front of the inlet, but rather a larger build-up inside the inlet, and, therefore, the recorded force values are not directly proportional to the mass increase. A qualitative comparison of the surcharges with no screen and a 2 mm screen at 1 cm digging depth can be seen in Figure 12. The use of vibration facilitates the leading edge cutting through the soil, with a force reduction of up to 17 %. When comparing the runs with 1 mm and 2 mm sieving plates in front of the inlet, respectively, it becomes apparent that the required forces are relatively similar despite the different screens. Even though the values for the 2 mm mesh in Table 4 are slightly higher for most runs, this is usually less than 1 N, which in the present context, is insignificant. This similarity in forces can be assessed positively here, as a finer mesh does not seem to produce more surcharge. This will, of course change if one would try to pre-separate with a very small aperture size (i.e., sub-millimetre) for the considered particle sizes; this, however, does not align with the purpose of this mechanism and its operating principle. Generally, it can be seen that vibrations reduce the overall necessary excavation forces with observed reductions up to 30 %; this result is in good agreement with the results reported by the authors before Just et al. (2021). However, one should keep in mind that with the very low forces required for the shallow digging depth (below 1.5 N), even heterogeneities in the analogue substrate being excavated can cause a significant difference in overall force requirements. For example, agglutinated particles, rock fragments, or a locally compacted area of increased relative soil density. Overall, the obtained results prove the viability of the mechanism and demonstrate its applicability to the presented objectives.
A concern which is prevalent when using sieves is blinding or clogging of screens, subsequently reducing screening efficiency and increasing residence times. Due to the, for terrestrial sieves, relatively large aperture size and the operating procedure (Section 3.2) no significant impact of this phenomena on the separation performance has been observed. Tests during the standardised sieving test have shown an increase in residence time when the sieve is not cleaned in between runs, however, the vibration motor in combination with the possibility of tilting the mechanism downwards offers a way of removing stuck particles. A shearing mechanism or bursts of compressed gas could further facilitate this (Section 6) (Singh, 2004). Table 5 shows an overview of the mass and preliminary power budget of the mechanism. The mass is differentiated between the mechanism mass and the mass of the camera with its moving platform, since the camera assembly mass is optional and mainly depending on the used lens, which is readily exchangeable. Even though the camera increases the capabilities of the mechanism, it is not required for the successful operation of LES 3 . Peak power consumption occurs when all stepper motors are moving simultaneously, and the vibrating motor is in operation. Idle power consumption refers to the steppers merely holding their position.
As mentioned in Section 2, LES 3 must be considered as a mechanism designed for the LUVMI-X rover platform in the support of ISRU applications, where contamination of the material collected from different sampling sites is not too problematic. This application is intended by design and the support of analytical applications, such as the use of mass spectrometers, was not the driver for the F I G U R E 12 Qualitative comparison of the surcharge at 1 cm digging depth between no front screen (left) and a 2 mm front screen (right). Dashed lines indicate the surcharge and show that the surcharge length in front of the inlet is shorter for no screen, but that material advances further into the mechanism in this case JUST ET AL.
| 275 development. During the laboratory test campaign, material got trapped in several locations which led to an overall material loss of 3-4%, but did not cause any operational detractors. Some possible changes to reduce this lost fraction can be found in Section 6.

| SUMMARY
The presented work introduces a mechanism capable of combining regolith excavation with its beneficiation in the form of grain size separation. Since LES 3 is intended for use with a small and lightweight lunar rover, LUVMI-X (Garcet et al., 2019;Losekamm et al., 2021), minimising the required excavation forces is crucial.
Laboratory tests in an analogue testbed have validated the working principle of the system and have shown that the mechanism, which mass is below 2 kg, is capable of excavating up to 100 g of soil in a single scoop while keeping the excavation forces at its maximum excavation depth of 3 cm below 8 N-around 40% of the traction force provided by the rover under lunar conditions. To facilitate size separation, LES 3 utilises a two-stage vibrating screen system. To investigate the feasibility of this system, an additional stand-alone dry sieving experiment was performed (Section 4) to identify a realistic separation level and to inform the operating requirements of gravitational vibrating size separation of regolith. It was shown that residence times scale with a power law when the aperture size is reduced. Thus, decreasing the necessary maximum particle size must be well justified. Furthermore, it is shown that batch sizes during vibrational sieving should be kept low (or sieve area maximized), as additional surcharge inside the sieve hinders the movement of particles and thus decreases separation efficiency. It is acknowledged that such size separation techniques can become challenging when applied in the low-gravity environment of the Moon and that the soil characteristics of regolith, especially its cohesion (Mitchell & Houston, 1972), pose another challenge; challenges which cannot be simulated here on Earth, as even the best simulants do not resemble the lunar regolith closely enough in all characteristics (Taylor et al., 2016). However, where a rough separation of particles by size is required and cross-contamination of sampling sites is not problematic, the presented system offers a simple way of achieving excavation and regolith grain size separation. We, therefore, are able to propose an alternative approach that eliminates an intermediate re- golith conveying step and is significantly simpler than more advanced separation methods like electrostatic or magnetic separation (Higashiyama & Asano, 1998;Rasera et al., 2020;Trigwell, Lane, et al. 2013). As many of the current ISRU processes rely on or benefit from a more controlled regolith feedstock (see Section 1), any level of particle size control will improve product quality and reduce the risk of ISRU process failure. Validation of the mechanism in a lunar gravity environment in the form of discrete element method (DEM) simulations, as well as how the sieving performance changes with reduced gravity, is subject of future work.
To further improve the capabilities of the mechanism, the following areas should be addressed and improved: • Dustproofing: Currently, the mechanism has exposed actuators, sensors, and optical surfaces. For application on the lunar surface, the stepper motors, where space-certified models must be implemented, must be dust-proofed to minimise the risk of failure and the implementation of mitigation techniques against the highly electrostatic properties of regolith must be considered. The same is valid for the motors of the pan/tilt platform on which the camera sits, which lens also must be protected from a dust cover of the regolith. An enclosure for the whole cam/follower system seems feasible, which would eliminate the risk of a jammed slider preventing the discharge of material.
• Material selection: As mentioned above, the LES 3 lab prototype is manufactured from aluminum due to its ease of machining and low financial implications, with some parts having been anodized to increase wear resistance. However, for a future version of LES 3 , different parts of the excavator will be manufactured from different materials. Whereas the inlet and the sieving plates must be made from a strong and wear-resistant material, other parts, such as the sliding doors, must be manufactured from a material with low sliding friction against the inlet material. In general, the use of coatings to reduce wear and keep granular material from sticking to surfaces seems beneficial and should be explored. The deployment of materials in a space environment and the connected challenges (e.g., radiation, thermal environment) must be considered and the space verification of all materials ensured.
• Sensor/actuator selection: The utilised vibrating motor, which facilitates both excavation and size separation, will be housed within the inlet structure at a suitable location. To maximise the scientific use of the mechanism, additional sensors could easily be integrated. This includes a loadcell to record the forces experienced by the inlet, a flow sensor to determine the flow rate of material inside the arm, and potentially an agitation system to increase sieving efficiency (as discussed in Section 4). This could, for instance, be a rotating paddle, but also small bursts of a compressed gas, where the latter could also be used to clean the sieving plates.
• Mass reduction and dead volume: To reduce launch mass as much as possible and free up maximum payload capacity on the rover, the overall mass of the system must be further reduced, which will require FEM analyses of all parts and subsequent mass T A B L E 5 Overview of the mass and power budget of the proposed mechanism. minimization. This is partly dependent on the mass of excavated and processed material required by the ISRU processes that are being supported in the given mission scenario, since this determines the scale of the inlet as well as the arm. Areas that can accumulate soil during operation must be minimized by design and mitigation methods implemented.
• Test on the rover platform at analogue site and with icy regoliths: To further evaluate the applicability of this prototype and due to the proposed polar deployment zone of LUVMI-X (Garcet et al., 2019;Losekamm et al., 2021), tests with icy regoliths or analogues prepared to comparable strength should be performed (Gertsch et al., 2006(Gertsch et al., , 2008Pitcher et al., 2016). Additionally, the mechanism is intended for a field analogue test with the rover in late 2021.
In summary, it was shown that the development of a dedicated excavation mechanism, targeting the readily powdered regolith top layer, with size separation capabilities for a lightweight lunar rover seems feasible. With an available traction force on the lunar surface of only around 20 N, engineering the mechanism for minimum excavation forces was crucial and implemented successfully. With several agencies as well as an increasing number of private companies currently developing small robotic vehicles for lunar applications, the design approach demonstrated in this study is applicable to the future development of excavation and beneficiation systems in the 100 s of grams range and emphasizes the need for specialized lunar excavation mechanisms once more. It also shows that, while being a discrete excavation method, low excavation forces are achievable during bulldozing motions, if the mechanism is specifically designed for it. The presented system can be utilised while the rover is moving and does not require the vehicle to stop (like other excavation mechanism; for instance, back hoes). The development of LES 3 also shows that, where powdered bulk regolith is required, focusing excavation activities to the top centimetres of soil simplifies the process due to limited effects of the changing regolith properties with depth.

ACKNOWLEDGMENTS
We appreciate the collaboration with Space Applications Services,