Regolith Advanced Surface Systems Operations Robot (RASSOR)

Regolith is abundant on extra-terrestrial surfaces and is the source of many resources such as oxygen, hydrogen, titanium, aluminum, iron, silica and other valuable materials, which can be used to make rocket propellant, consumables for life support, radiation protection barrier shields, landing pads, blast protection berms, roads, habitats and other structures and devices. Recent data from the Moon also indicates that there are substantial deposits of water ice in permanently shadowed crater regions and possibly under an over burden of regolith. The key to being able to use this regolith and acquire the resources, is being able to manipulate it with robotic excavation and hauling machinery that can survive and operate in these very extreme extra-terrestrial surface environments.


INTRODUCTION
Space resource utilization promises to revolutionize the human condition [I], but the first step in this activity is acquiring the space resources to be used as feedstock . In this case the space resource under consideration is extraterrestrial regolith, with emphasis on lunar regolith, since it is well characterized from previous missions such as NASA ' s Apollo program. The lunar regolith consists of approximately 42% oxygen by mass, [2] which can be extracted using chemical engineering methods and then used in space for rocket engine propellant, life support air, water production and usage as a consumable gas [3] . In addition, there are other volatiles present in the lunar regolith such as water, hydrogen, helium, carbon monoxide and helium 3 which are all potentially valuable in a space resource utilization system [4]. Other byproducts of extracting these volatiles are useful metals such as aluminum, iron and titanium.

REGOLITH EXCAVATION NEEDS
Since typically more than 90% of the mass required for chemical rocket propulsion systems (such as hydrogenoxygen combustion engines) is propellant [5] , the highest priority in space resource utilization is to make and use propellants in-situ. Large mission mass savings will result since the deep gravity well of the Earth, which accounts for 9.3 -10 kmls of delta-velocity to reach low Earth orbit from Kennedy Space Center at an orbital inclination of 28 degrees, can be avoided . The logistics train can also be simplified, and the robustness of a mission can be improved as the crew becomes more self-reliant, without regular critical supply shipments from Earth.
If two NASA Constellation class lunar lander missions are assumed per year, then the oxygen propellant required for a chemical propulsion system is approximately 4 tonnes for a four crew, human class vehicle, resulting in 8 tonnes of oxygen required per year. In addition, 2 tons of oxygen would be required per year for life support and fuel cell consumables. Therefore, the regolith required would be proportional to these needs, since the oxygen would be extracted from the regolith. The exact quantities are discussed below and vary based on the chemical process used (I %-28% "efficiencies) and the possible presence of water ice (5%-10%) at the lunar poles [ 4] The Regolith Advanced Surface Systems Operations Robot (RASSOR) project assumed that the near-term miSSions would be robotic precursor landers with limited total payload masses of fewer than 500 kg. These robotic precursors will prove that regolith excavation and utilization is possible as a technology demonstration . Subsequently, multiple micro-excavators operating in a swarm can be delivered on small landers and meet total mission requirements in a scalable and affordable fashion.
In this paper a prototype robotic micro-excavator is presented that can meet the requirement of excavating enough regolith feedstock to supply 2.5 tonnes of oxygen per year. To achieve the I 0 tonnes of oxygen, four RASSOR micro-excavators would be deployed and operated simultaneously.

RASSOR CONCEPT OF OPERATIONS
The nominal mission of RASSOR is a five-year long mining operation to deliver lunar regolith to an oxygen production plant. (The plant itself is outside the scope of this project.) An estimated 255,500 kg of regolith needs to be delivered in order for a plant of approximately I% efficiency to produce 2,555 kg of oxygen in that period of time.
RASSOR will arrive on a lander comparable in size to the Mars Phoenix Lander and deploy itself from an approximately 1-meter-high deck by simply driving off the edge, which sidesteps the need for ramps or other offioading systems. RASSOR is designed to withstand the impact of the fall , land on any side and right itself if necessary. The excavator will need to drive more quickly than previous space exploration robots in order to meet its mining requirements, so it will be free of hazard avoidance software, which would slow it down. Instead, the landing site will be selected in an area that is as flat as possible, and RASSOR will perform "acrobatics" (the process of using the arms to assume various positions) to handle the obstructions and loose soil it will encounter. It will drive over small rocks and either drive around rocks or use acrobatics to pull itself over larger rocks, furrows or ridgelines. In the event of traction loss, RASSOR can somersault to become unstuck or rise up on its arms and roll on the bucket drums, then lower itself and resume normal driving.
RASSOR will drive at least 100 meters away from the lander to avoid kicking dust up onto the lander' s solar panels during mining operations and to provide ample U.S. Government work not protected by U.S copyright surface regolith. The round trip takes 17 minutes, with RASSOR driving at a speed of at least 20 cm/s. In comparison, the Mars Science Lab rover has a top speed of 4 cm/s. It will collect 20 kg of regolith each trip, making a total of 12,775 trips per year to meet regolith mining requirements. This amounts to 35 trips per day (24-hour period). If the mobility portion takes 17 minutes and the mining takes 10 minutes, then a 30 minute mining cycle will suffice, and the RASSOR can operate for 16 hours per day to achieve its goal of 255 ,500 kg of regolith mined. It can mine the upper surface layer, or it can trench to access water ice that may be located a meter or more below the surface. The capability of digging in areas with higher concentrations of ice is being tested and will also possibly have a higher yield of 5.5% water. This will require that RASSOR take very small scoops-more of a shaving action -and would be a much longer operation than digging in drier surface regolith. However, much depends on the characteristics and properties of the icy regolith, which are unknown at this time.
At the end of each mining trip, RASSOR returns to the oxygen production plant near the lander, where it raises itself into a vertical dumping position and reverses rotation of its bucket drums to dump the regolith into a receiving hopper that can be up to a meter high off the ground [6]. RASSOR is powered by batteries that will be recharged at the lander in between mining treks as required . The concept of operations has factored in 8 hours per Earth day for recharging operations.

RASSOR GENERATION I REQUIREMENTS
After the needs and concept of operations had been identified the more detailed requirements were written. Some of these requirements are specific about how they shall be met, although, a number of the requirements are meant to bound the problem without forcing a specific design . This is to allow the team to be more creative while still meeting requirements to achieve to overall mission. These bounding requirements fit within the Technology readiness level (TRL) 3: " Analytical and experimental critical function and/or characteristic proof of concept." [7] • RASSOR shall successfully deploy itself from the lander.
• RASSOR shall drive I 00 m, excavate, and return to the lander.
• RASSOR shall have a maximum mass of 50 kg, with a preferred mass of 20kg or less.
• RASSOR shall mine the top 5 em of surface regolith for nominal regolith mining operations • RASSOR shall be capable of mining I meter deep using a slot dozing trench method , for icy regolith mining and science observations .• •.
• RA SSOR shall successfull y mine 700 kg of rego lith within 24 ho urs.
• RA SSOR shall be equipped with o ne or mo re cameras.
• RAS SOR shall recharge its battery at the land er usi ng a dust to lerant connector.
• RASSOR sha ll have a minimum lifespan of 5 years.
• RASSOR shall have the ability to self-ri ght itself.
• RA SSOR shall be tele-operated whil e it will still offer the o ptio n to later add redundancy and autono my for th e dri ving system.

RASSOR GEN ERATION 1: DESIGN SOLUTION
RASSOR uses two bucket drum excavato rs to acco mpli sh its primary fun cti o n of co llectin g and transporting rego lith . A bucket drum excavator is a novel device that excavates, stores and dumps rego lith . Lockh eed Martin Space Systems, Denve r (under a contract fro m NASA) develo ped the firs t bucket drum fo r use on a small roboti c excavator [8]. T he Granular Mechani cs & Regolith O perati ons (G MRO) La b at ASA Kenn edy Space Center has deve loped a nove l appli cati on of th e bucket drum concept by employ ing two counter rotatin g buc ket drum s in a roboti c pl atform that has advanced pos itio ning and posing capabilities.
T he main adva ntages of a bucket drum are th at the excavatio n scoops are sma ll and staggered so that at any given tim e onl y o ne or two are engaged in the rego lith , thereby keeping th e excavatio n forces low, and th e regolith collected becomes trapped inside the drum due to a set of baffl es until the directi on of rotati on is reversed. Thi s approach compl ete ly eliminates the need to ha ve a separate regolith storage and dump bin. The bucket drums used o n RA SSOR each consist of 5 segments with 3 scoops per segment (see fig ure I). Each bucket drum is designed to co ll ect and ho ld I 0 kg of regolith when 60% full . T hey have scoop openings that can accommodate rocks up to 5 e m in di ameter. T hese rocks may be excluded thro ugh a gratin g o r oth er system in the future.
Aluminum sheet metal was used to construct th e scoops and baffles whi ch were placed aro und pl ates that di vide the segments. Five a luminum rods with tapped ends run th e length of th e bucket drum and attac h to the end-caps, compressing the segments together. T he scoops on the bucket drums a lso have sta inl ess steel removable cutting edges th at are ri veted o n. T he sheet meta l approach kept th e mass of the bucket drums to about 7 kg each. T he he ight of the bucket drum s with respect to th e ground plane needs to be actu ated in order to engage the soil and contro l th e cutting depth of the scoops. RASSOR accompli shes thi s by placing the bucket drums on th e ends of sing le degree-of-freedom arms (see fi g ure 2). Each arm is driven at a rotati onal jo int on the RA SSOR chassis and houses the motor and gearbox fo r th e bucket drum rotatio n. T he bucket drum rotati on motor and gearbox are mo unted inside the structure of the arm , and a drive shaft transfers th e torque fro m the gearbox o utput to a bevel gear pair, whi ch in turn drives the bucket drum (see fi g ure 3). Inside th e RASSOR chass is, a motor, pl anetary gearbox, and wo rm gear system is used to dri ve the arm rotation. The arms also were des igned to perform acrobatics as a useful mobili ty fea ture. Usi ng the dual arm confi gurati on, th e ve hicle is abl e to rig ht itself if flipped over, climb over obstac les much ta ller than th e tread height, dump regolith at heights taller than o ne arm length, and stand th e chassis up to clean out debri s from th e tracks (see figure 4).
T he mo bility system o n RASSOR was a lso des igned to be very si mpl e as the primary goal of thi s version is to prove the concept of low reacti on force dual bucket drum excavati o n. To th at end , th e mobility system is a tank tread design that was initially modeled from similarl y-sized commerciall y ava ilabl e drive train s. During testing, however, it was observed that th e fin e particles would get caught between the treads and th e drive pulleys, causing the treads to jam up or track off the pull eys. To fix thi s, new drive pull eys were designed that have large openings between the teeth so that the so il can c lear out fro m treads.   T he crawler-way fin es compacted we ll , whi c h provided a way to test RASSOR' s ability to overco me hi gher excavati o n fo rces. RA SSOR was abl e to di g successfull y, as long as the di gg ing depth was very shallow (less than a centimeter), whi ch meant it took much longer to fill th e bucket drums. The crawlerway fin es a lso conta ined a lot of rocks and grave l, whi ch tended to get caught between th e whee ls and the tracks and ca used much trouble during testing, espec iall y during counter-steerin g. (See fi g ure 12 in th e Lessons Learn ed secti on). RASSO R a lso demonstrated th at it was able to cl imb a 20 degree slope, turn , and drive latera ll y. While it fa iled to c limb stra ight up a steeper 30 degree slope, thi s was because th e crawler fin es we re in a loosely piled mo und th at sheared and ava lanched under RASSO R' s we ight and caused it to slide backwa rd s. It is expected that RA SSO R wo uld have succeeded if th e hill U.S. Governm ent work not protected by U.S. copyri ght 5 was more co mpact, and furth er testing will be perfo rm ed to resolve thi s. In addition, th e bucket drum scoops may be used as a "climbing pito n" device used fo r contro ll ed ascending and descending, and thi s hypothesis will a lso be tested in the future. The BP-I simul ant had th e hi ghest lunar simul ant fid e lity but was o nl y available in one small outdoor test bin . RA SSO R was able to successfull y dri ve, perfor m acro bati cs, di g loose surface rego lith, and trench. Measurements were taken that confirmed th e bucket drums were able to co ll ect not o nl y I 0 kg of regolith each, as des igned , but upward s of 14.5 kg. Current draw was measured and recorded for di fferent acti vities (see fi g. 8.). RASSO R pulled 3 to 5 amps whil e dri v ing; 8 to I 0 amps while ra ising its chass is into the iron cross positi on; 3 to 4 amps while lowering th e chassis bac k down; and 2 amps during shallow mining. Temperature was also measured and exceeded 150 degrees F after less th an two hours of o utdoors testin g (see fi g. 9). Thi s was of concern because th e motor contro llers have a rated maximum operating temperature of 180 degrees F, so temperature will continue to be monitored, parti cul arl y when testin g o utdoors in th e heat. Future versions of RA SSOR will include provisions fo r cooling and will provide traceabil ity to lunar conditio ns with a suitabl y sized radi ato r on the top surface.
RASSOR was a lso tested in BP-I using a setup that offloaded 5/6'h of its weight to simul ate lun ar grav ity. O ne bucket drum was lowered into th e simulant to mine, and the excavati on reaction forces were greater than th e tracti o n, w hi ch translated th e ve hi c le witho ut excavating so il. When th e second drum was lowered, RASSOR sto pped sliding and both drums began to coll ect so il , thereby demo nstratin g the effi cacy of dual counter-rotatin g bucket drum system. It a lso appeared that verti cal reactio n fo rces were low eno ug h not to influence RA SSOR' s di gging, and th ere are futu re test pl ans to quant ify them.
RASSOR was tested again w ith th e gravity off-loader, th is time in icy rego lith, whi ch was a mixture of BP-I with I 0% water by weight, th at was compacted and cryogeni call y frozen to 83K using liquid nitrogen.
RASSOR demonstrated it was able to mine, wi th the advantage that frozen BP-I was very brittle and ab le to be broken up . A shortcoming of the test, however, was that it was performed outside in the summer, where the air temperature exceeded 90 degrees F and continually warmed the test bed. Prior lab tests showed the BP-1 stayed frozen at the core for an hour. But during testing it appeared that the outer layer was easier for RASSOR to scrape off until it hit a harder, more frozen layer, which then subsequently warmed. For future tests, it wo uld be desirable to have liquid nitrogen keeping the regolith frozen during testing, whi ch would involve a more complex test setup but have higher fid elity. Run Time (m in) Figure 9 -RASSOR internal temperature during two hours of te ting. Sharp drops in temperature correspond to shutting RASSOR off for periods of time, during which data was not recorded.

LESSONS L EARNED AN D GOOD PRACT ICES
Improvements to the des ign of RASSO R were found during fabricatio n, assemb ly, and testing. The three sections below summarize the lessons learned during these stages.

Fabrication
One improvement learned from the fabrication process is to des ign for minimized part count rather than ease of fa bricati o n. In many cases, it may actuall y be qui cker to machine a co mplicated part on a multi -axis CNC rather th an machine and asse mble severa l parts o n a simpler mac hine, such as a wa ter-j et cutter. This wo uld also avoid problems later on during asse mbl y, as those multipl e parts may need to be alig ned very closely to function properl y. A sing le piece part avoi ds stacked to lerances.
Assembly W ith a ti ght requirement on size and weight, it was benefi c ial to pack hard ware in to the chassis as ti ghtl y as possible. T hi s resulted in limiting the accessibili ty of many components during both asse mbl y and maintenance. As more co mponents of the system were tested, it was necessary to remove some o f th em to fi x or adjust certain aspects. If th e who le ro bot had to be d isassembl ed to get to the drive wheel motor, fo r example, it became very time consuming and impracti ca l. Another probl em with takin g components apart and putting them back together many times is th at so me may actua ll y wear o ut in ways th at wo uldn ' t happen du ring norm al robot operations.
Anoth er po int is to ensure th at wi rin g has bee n wellacco unted fo r. Des igners will o ften create CAD mode ls th at neglect the wires connecting all th e motors, contro llers, cameras, etc., whi ch makes it hard to envision th e fin al setup . W ires may run very close to o pen gears and shafts and could get caught as the rover moves ove r rough terrain . Additio nall y, wires may block access to components th at need to be adjusted in place, such as th e tensio ner fo r a cha in or be lt. In the case of RA SSOR, these pro bl ems we re miti gated after fin a l asse mbl y of the robot, but in th e future , including th em in the model wo uld provi de better res ults .

Testing
With a compl ex des ign, the testing phase will typi call y reveal the most areas in need of improve ment. T his was certa inl y th e case with RASSO R.
The first subsystem th at requi red many adj ustments were th e trac ks. It turned out to be very chall eng ing to design a track dri ve system that could work in a vari ety of soi ls analogous to lunar regolith . Ini tia l tests in sand exposed the c logging problem, where regolith particles accumul ated between th e whee ls and the trac ks, causing th e trac k to lose tooth engagement and slip off the whee l. Thi s was remedi ed by redesigning o pen whee ls that allowed the dirt to fl ow out toward s th e hub as the wheels turned. Thi s so lved th e clogging issue for most types of soft a nd hard soi l, except fo r those like th e craw lerway fi nes with larger size rocks th at could still get wedged in the wheel (fig 12).
U.S. Government work not protected by U.S. copyri ght 7 Figure 12 -Crawler-way gravel lodged in the track, which pushed up against the arm crossbar and locked up the drive system.
Another improvement was to exchange an all-rubber be lt materi al with one th at enco mpassed stainl ess steel wire ro pes running continuo usly th ro ugh it . T his corrected a belt stretch issue that had been contributing to the belt co ming off th e wheels when clogged. That iss ue was exaggerated at elevated temperatures due to th e rubber softening. It sho uld be po inted o ut the on th e moon the oppos ite iss ue wo uld occur where the rubber wo uld lose its fl ex ibility. However, rubber wo uld not be used o n the moon; but as a terres trial analog it was acceptabl e in thi s prototype.
Initi al testing o f the arm s revea led th at the cha in connectin g two gearboxes in series worked as designed in one directio n, but skipped teeth under load in th e other directi o n. Thi s was due to th e entrance ang le of the chain o nto the dri ve sproc ket not being equal in both d irecti ons. T he iss ue was so lved by ti ghtening th e cha in , whi ch had to be do ne repeatedl y, as the chain stretches. With thi s improvement in pl ace, the next issue that occurred was too much fri cti on between gears on a c ustom gearbox ho using using off-theshelf gears. T he root of thi s probl em was a mi sali gnment between th e gears caused by to lerance stac kin g among the multiple parts comprisi ng th e gearbox ho using.
A genera l improve ment to be made during the testing phase is to test subsystems that appear multiple times on a sing le unit prior to integrati o n. This will a llow adjustments to be made to th e single assembl y rath er th an all copies at o nce. T hi s was most apparent on th e arm mechanism, whi ch utili zed fo ur identi cal gear tra ins that underwe nt several iterations. T he speed at whi ch iterati ons were made wo uld have increased if updates had been made to a sin gle uni t on ly.
A set of encoders was linked to th e output of the moto r thro ugh a set of spur gears. The back las h of th e spur gea rs coupled with some wo bble in th e assoc iated ba ll bearings was eno ugh to confuse th e moto r contro llers and cause the moto rs to stop intermittentl y. T he last set of encoders was attached di rectly to th e o utput shaft of the motor, whi ch was o ptimal. However, even thi s setup needed o ne improvement primarily because of the way the encoder wheel was designed . The plastic encoder wheel was intended to simply press-fit onto a shaft, which in RASSOR' s case was smooth and had high accelerations and decelerations. This resulted in the encoder coming loose. Knurling the surface of the shaft and using an adhesive solved this negative. It is highly recommended to use encoders with positive locking encoder wheels directly on the motor output to avoid this issue .
The remaining lessons learned do not suggest hardware changes, but rather refinements and proper implementation of the operations concept. The original concept of the counter-rotating bucket drums was to make shallow skim cuts off the soil surface and drive while digging to fill up the scoops. In practice, this turned out to be difficult due to the purely manual control that required constant driver input. If there was any initial unevenness in the soil surface, those bumps would make the rover move off level as it drove forward , which meant one drum would be raised off the ground while the other would be driven deeper into the soil. This created more unevenness for the next time the rover would make another pass. Another unintended side effect of the deep cut is that under some soil conditions the scoops could get clogged with compacted regolith, and therefore become ineffective. On the moon, low gravity, electrostatic forces, Van der Waal forces, and high friction forces between the particles could create a similar situation whereby cohesion of the granular material is increased , causing similar bridging and clogging. This will be addressed in future designs by opening the size of the scoop opening to prevent bridging or by vibrating the drum to free the particles. The digging depth problems could be mitigated by automating (with scripting) the skim cut operations so that the driver would not need to try to adjust the arms constantly to keep them level even as the rover chassis bounced. Driving very slowly while digging would allow the arms to keep up with the moving chassis in order to take even, shallow cuts. The same scripting approach should be taken for the acrobatics moves in order to smooth out the loads during those maneuvers .
U.S. Government work not protected by U.S copyright 8 Lastly, testing in icy regolith required some changes in the digging approach. It seems to be more effective to first break up the hard, icy soil and then scoop up what has been broken loose. A faster drum rotation will help the breaking up step, while the slower drum rotation will then be used to pick up the soil. Improvements in the cutting edge of each scoop, such as serrations or sharp pick ahead of the cutting edge will also help the rover perform better while digging icy regolith.

SUMMARY
A novel, compact and lightweight excavation robot prototype for manipulating, excavating, acquiring, hauling and dumping regolith on extra-terrestrial surfaces has been developed and tested at NASA, Kennedy Space Center. Lessons learned and test results have been presented in this paper, including results from digging in a variety of lunar regolith simulant conditions as well as frozen regolith mixed with water ice.
This prototyping effort has shown prom1smg results and proven the concept of using counter rotating bucket drums as an effective method of manipulating regolith in a load, haul and dump scenario, to produce a micro-excavator system that can be delivered to the moon and other extraterrestrial bodies on small robotic landers. This method successfully mitigates the problem of only having low digging reaction forces available in low gravity environments, which is a major challenge when using traditional excavation methods, such as those used on Earth .
The lessons learned have been valuable and the testing has also revealed opportunities for improving the design and operations. A second generation RASSOR will be designed, fabricated and tested to take advantage of these opportunities. Eventually, it is hoped that a swarm of RASSOR ' s will operate on the moon and other extraterrestrial bodies to enable regolith mining for space resource utilization.