RASSOR

NASA/Frank Michaux. Public domain (NASA / US government work).

Overview

RASSOR is a regolith excavator developed at NASA KSC as a technology demonstrator for low-mass excavation in reduced gravity. Kennedy conceived the counter-rotating bucket drum method in 2011 and demonstrated a prototype at that year’s NASA Lunabotics Mining Competition [8]. The approach carries forward into the ISRU Pilot Excavator, a 30 kg-class machine intended to excavate and transport 10,000 kg of lunar regolith on the surface [1]. The method is protected as US patent 9,027,265, “Zero horizontal reaction force excavator”, granted 12 May 2015 [2].

Specifications

Parameter	RASSOR 2.0
System mass	66 kg, half of RASSOR 1.0
Regolith payload	80 kg, double RASSOR 1.0
Excavation throughput	2.7 t/day minimum, single unit
Mass per unit rate	0.38 kg per kg/h
Power per unit rate	4 W per kg/h
Bus voltage	48 VDC
Average driving velocity	27 cm/s
Maximum driving velocity	56.5 cm/s
Actuator count	10

Values from [2].

RASSOR 3.0 is intended to reach TRL 5, and thermal control in vacuum and a battery able to survive extreme temperature excursions are the identified remaining problems [2].

The NASA technology transfer page for the machine adds four properties the conference papers do not state: drum rotation is about 20 revolutions per minute during excavation, the chassis is symmetrical so that the machine continues to operate after overturning, the arms fold to reduce launch volume, and control is wireless with onboard transmitting cameras [10]. That page lists space construction alongside in-situ resource utilization, regolith and water ice mining, and terrestrial mining in hazardous locations as intended applications, which is the basis for classifying the machine as construction rather than as a drill.

No flight has been assigned.

Excavation mechanics in low gravity

Terrestrial excavators oppose digging reaction with weight transmitted to the ground through wheels or tracks, so machine mass scales with excavation force. A mobility platform must therefore either be massive enough to overcome the reduced-gravity weight effect or reduce the digging forces themselves, and since space missions are mass constrained the second is the cheaper route [8]. Launch cost is of order 8800 EUR per kilogram depending on the vehicle [7]. A shovel or drill on a light chassis in one sixth gravity pushes the machine away from the surface rather than cutting.

The material is not weak. In-situ lunar soil relative density is about 65 percent in the top 15 cm and exceeds 90 percent below 30 cm, against 65 to 75 percent as the practical limit of terrestrial field compaction with heavy equipment [5]. Regolith bulk density ranges from 1500 kg/m3 in highland samples to 1640 kg/m3 in mare samples [4]. Shear strength rises with relative density, so digging forces remain substantial where weight does not.

Force reduction has been attempted several ways. Pneumatic mining achieved 8000 g of regolith excavated per gram of transport gas, at the cost of a consumable [8]. Percussive excavation fluidizes dry regolith ahead of the tool edge by separating interlocking grains: a 30 inch bucket driven at impact energies of 13.6 to 30.5 J and 0 to 700 blows per minute reduced the average maximum excavation load by up to about 50 percent, against about 70 percent reported for a much smaller Honeybee Robotics scoop at 2.5 J and 1000 blows per minute [6], and work by Kennedy with Honeybee Robotics and Berkeley reported reductions as large as 90 percent in simulant. Shallow-angle scraping plates, bucket chains, bucket ladders and bucket wheels take many small cuts and require a separate hauling stage.

Lockheed Martin stacked several bucket wheels into a bucket drum in 2008 and 2009, which also served as the container for excavated material; field testing at Mauna Kea showed the drum effective but still reliant on traction for excavation and mobility [8]. Counter-rotation removes that dependency by canceling the horizontal reaction rather than reducing it.

Approach

RASSOR carries a bucket drum on each end of two opposed articulated arms. Bucket drums are hollow cylinders with scoops staggered around the perimeter; regolith enters through the scoops, slides down an internal baffle system that prevents it falling back out, is transported inside the drum, and is discharged by reversing rotation so the material slides back out through the same openings [1]. The drums on opposite arms dig simultaneously in opposing directions, so the horizontal excavation reactions cancel and the net force transmitted to the chassis approaches zero. Excavation force is reacted internally through the structure rather than externally through traction. Bucket drums were developed by Lockheed Martin in 2008 and have been used on robotic excavator prototypes since [8]. Counter-rotation cancels the horizontal components while the vertical components remain small, so the system is not primarily reliant on weight or traction [7]. Mirroring the two drum sets across the chassis also cancels the axial forces, and placing the scoops at the outboard edges pushes overflow material clear of the vehicle instead of piling it in front.

RASSOR’s drum is the large size in the Kennedy scaling series.

Drum	Diameter	Width	Scoop width	Scoop height	Mean fill per drum
Small	237.5 mm	206 mm	51.6 mm	26.4 mm	3.80 kg
Medium	295.1 mm	254 mm	63.5 mm	34.5 mm	7.30 kg
Large (RASSOR)	437.1 mm	358 mm	90.4 mm	47.8 mm	24.98 kg

Values from gantry testing in BP-1 simulant [1]. A four-drum machine using the large drum holds 99.94 kg.

Each drum is divided into four sections carrying two scoops each, giving eight buckets per drum [2]. Construction is aluminum sides, carbon fiber scoops and baffles, and stainless steel cutting teeth [1].

The drum design envelope was published in March 2020 as the NASA RASSOR Bucket Drum Design Challenge on GrabCAD, which drew approximately 350 entries against a requirement to maximize regolith extraction and retention while remaining free of obstructions during loading and unloading [7]. The stated limits were 175 mm maximum total scoop width engaged at any instant, 5 kg maximum drum mass, 450 mm maximum drum diameter, 360 mm maximum drum length, and 17.6 liters minimum captured volume. The fourth-placed entry, a single 3D-printable part combining an outer double helix, a 180 degree spiral tunnel behind each of two scoops, and an inner double helix, massed 4.93 kg at a fill ratio of about 75 percent [7].

Scoop geometry sets dig depth and flow rate into the drum, and scoop height sets the effective internal diameter and hence total capacity [1]. Digging at the full scoop depth lets regolith bridge across the opening and reduces the mass collected per rotation; several hundred hours of RASSOR 2.0 testing established that limiting cut depth to at most 50 percent of the scoop opening collects more regolith per rotation. Gantry tests [1] at cut depths of 10 mm and 40 percent of scoop height, linear cut speeds of 10 and 30 mm/s, and a drum tangential velocity fixed at 8.5 times the linear cut speed to hold cut pitch constant, showed horizontal excavation force per unit drum width rising with both cut speed and cut depth, and mechanical excavation energy per kilogram collected falling with increasing cut speed, increasing drum size, and decreasing cut depth. Lateral force was negligible throughout, and the vertical force once the drum was fully engaged equalled the weight of the captured regolith.

The arms raise, lower and rotate the chassis over the drums, so the vehicle can right itself if overturned, and can fold into a Z configuration with the chassis vertical, one set of drums and wheels on the ground and the other reaching out from the top of the chassis [2].

Mobility

RASSOR 1.5 used tracks, and metallic link tracks in Black Point 1 basalt simulant required many design iterations to become reliable [2]. Because the counter-rotating drums cancel horizontal excavation force, a large drawbar pull is not required, which reopened the running gear choice. Free driving, drawbar and tilt table testing in BP-1 at 1 g compared metallic link tracks against 13 inch wheels in four and six wheel configurations and 17 inch wheels in a four wheel configuration [2]. On the 25 degree tilt table the 17 inch four wheel configuration succeeded in three attempts of three, where tracks and both 13 inch configurations each succeeded once in three. The 17 inch wheels drove comparably to tracks with slightly lower drawbar pull and fewer failure modes, and were selected [2]. The derivative Pilot Excavator did not carry the RASSOR wheel forward: its wheel started instead from a modified VIPER wheel with removable grousers, cleat surfaces and spokes, and side rings chamfered to act as a ski during skid steer turns [9].

Average driving velocity is 27 cm/s and maximum velocity 56.5 cm/s. Moving the bucket drums shifts the vehicle center of mass, which produces distinct driving modes used deliberately for operational versatility [2].

Extraction from soft ground uses the arms. If RASSOR becomes stuck, the drums are lowered until vehicle weight is shared between wheels and drums, reducing ground pressure enough to continue driving; loaded drums add weight where it is needed [2].

Power and energy

RASSOR 2.0 runs on a battery at a 48 VDC bus [2]. Actuator power draw measurements from the component test campaign sized total battery capacity; the capacity itself is not published. Motors matched to the required torque and speed lost continuous torque near rated speed at 48 V [2]. A bus above 100 V would have held continuous torque flat across the speed range but required a custom battery or series batteries, which schedule and funding did not permit. The adopted solution kept the 48 V bus and selected a motor winding holding continuous torque constant across the speed range at the cost of a higher maximum rated speed, with PI loop tuning recovering low-speed performance [2].

The operating concept assumes a 24 hour cycle of 16 hours mining and 8 hours recharging; at 27 cm/s average velocity, an 80 kg payload and a mining site 100 m from the lander, that yields about 35 mining trips per day and 1,000,000 kg of regolith per Earth-year, the quantity required to produce 10,000 kg of oxygen per year at an assumed 1 percent yield [2].

Thermal

RASSOR 2.0 is a laboratory prototype with no vacuum thermal control [2]. Radiating heat to space during the lunar day and conserving internal heat at night are named as the principal enhancements required for RASSOR 3.0, together with sourcing a battery that survives extreme temperature events [2]. Lunar surface temperature at 45 degrees latitude ranges from a mean 350 K at local noon to a mean 89 K before sunrise [4].

Compute and avionics

The onboard computer is an RTD CMX32MVD1860HR-2048 PC/104 form factor single board computer. Ten actuators are driven by Elmo Motion Control G-Sol WHI20/100 controllers, which read their feedback sensors directly and close proportional-integral loops for position and velocity, and the bucket drum and arm controllers run a dual loop of position over velocity [2]. Every actuator carries a US Digital EM1 optical quadrature incremental encoder at 10,000 counts per revolution plus three factory-fitted Hall effect sensors on the Parker Bayside motor for commutation, and the four bucket drum actuators and two shoulder actuators additionally carry Netzer DS-70 single turn absolute encoders at 19 bit resolution, read over SSI, with the actuator parameters below.

Actuator	Max torque	Continuous torque	Speed	Mass
Shoulder	644 N m	236 N m	~16 / ~10 rpm	3.58 kg
Bucket drum	191 N m	93 N m	~25 / ~18 rpm	1.30 kg
Drive	191 N m	93 N m	~25 / ~18 rpm	1.17 kg

All actuators use a 161:1 reduction, and the shoulder actuator is built around a Parker K089050 frameless motor with an SHG 32-160 harmonic drive component set and carries a safety brake so a loaded arm can be held without powering the motor [2].

System-level sensing is an Xsens MTI-30-2A5G4-O inertial measurement unit, forward and aft stereo camera pairs, and forward and aft hazard cameras, all cameras Axis Communications P1224-E network units [2]. On the gantry test stand used to characterize the drums, force and torque were logged at 1 kHz through an NI-9237 C series module while motor feedback was oversampled to align with it, the lower motor rate set by the CAN bus interface [1].

Autonomy

Software is built on ROS and split between the vehicle and a driver station [2]. The onboard side talks to motor controllers, sensors and cameras and publishes health and status.

Situational awareness comes from the stereo cameras fused with the IMU and actuator feedback, and the autonomy does not depend on high-accuracy sensing. Hazard cameras are used for inspection rather than navigation: quantifying wheel sinkage, tank-steer surcharge and wheel slip, and observing dig depth and regolith bridging as material enters the drum scoops [2].

The reference concept of operations is fully autonomous. RASSOR uses its arms to lower itself off the lander deck, avoiding a separate deployment mechanism, then drives to a mining site assumed to be 100 m away [2]. On the first traverse it drives slowly, surveying with stereo and hazard cameras and building a hazard map held onboard. Subsequent traverses use that stored map with low-fidelity stereo vision to recognize landmarks, including its own previous wheel tracks, and run faster. At the site the drums are lowered and excavation proceeds while driving slowly forward, with software controlling depth of cut and balancing excavation force between the two drum sets. Depth of cut is bounded by the bridging limit: cutting deeper than about 50 percent of the scoop opening lets regolith bridge the opening and reduces mass collected per rotation [1], while horizontal excavation force per unit drum width rises with both cut depth and cut speed. Slot trenching is possible because the drums extend beyond the wheelbase: the machine cuts a trench then drives into it to reach regolith deeper than 1.0 m, where higher water concentrations are expected. Torque sensing at the shoulder actuator determines when the drums are full, triggering return to the lander and discharge over a receiving hopper. Autonomous operation has been exercised in ROS and Gazebo simulation [2]. Drum load is also estimable without torque sensing: a free-spinning drum current model fitted to motor current reached about 7.4 percent mean percentage error on hardware [3].

Communications

The link between vehicle and driver station is wireless, using a Ubiquiti Networks PicoStation M2HP 2.4 GHz radio [2]. This is terrestrial test equipment; no flight communications architecture is defined at this technology readiness level. The driver station pulls camera streams over IP from the vehicle [2].

Modes of operation

Side	Mode	Behavior
Vehicle	Autonomous	Executes the concept of operations from onboard sensing, publishing health and status to the ground
Vehicle	Tele-operation	Waits for and executes streamed commands
Driver station	Supervisor	Accepts health, status and camera streams while the vehicle runs autonomously
Driver station	Tele-operation	Full command authority over the vehicle

Modes from [2].

Semi-autonomous behaviors can be invoked from within tele-operation. In Auto Dig the onboard computer drives straight at a commanded speed while balancing torque between the two bucket drums, so the operator sets speed and the vehicle maintains the force balance [2]. Force balance between the drum sets is what keeps the net horizontal reaction near zero [7].

Ground operations

Ground control is a single driver station, a Microsoft Surface Pro 3 with an Intel Core i7 and 512 GB of storage [2]. RASSOR is a laboratory and analog-field prototype operated by its engineering team at NASA KSC. No planning cycle, shift structure or flight operations tooling is defined.

Technologies developed

The counter-rotating bucket drum architecture, patented as a zero horizontal reaction force excavator [2], is carried forward by the ISRU Pilot Excavator, whose 30 kg-class design was sized against the Kennedy drum scaling dataset; that dataset concluded the small and medium drums collect enough regolith at the required rate for the IPEx concept of operations [1]. IPEx is being built to a TRL 6 flight-ready state for a notional technology demonstration mission that would excavate up to 10,000 kg of regolith at 42 kg/h and cover 70 km in the lunar south pole region over 11 days [9].

Onboard mass inferencing estimates how much regolith is in the drums during operation, which a closed-loop ISRU process needs in order to deliver a defined feedstock rate. Three approaches were formulated and two tested on hardware. An arm raise model infers mass from the electrical power integrated over an arm raise. A free-spinning drum current model infers it from motor current with the drum turning unloaded; on hardware it gave a mean percentage error of about 7.4 percent once two outliers were dropped, falling to about 2.6 percent for masses above 20 kg [3], and a neural network augmentation of the same model was integrated into an advanced robotic system. An excavation drum current model, inferring mass from current during digging, gave preliminary R squared of 0.76 [3]. Named next steps are building models from lunar or simulated low-gravity excavation data and exploring online and transfer learning, since a model fitted at 1 g in simulant cannot be assumed to transfer to one sixth gravity.

The actuator design is transferable in its own right: rotary units combining a frame-size-constrained motor, a 161:1 harmonic reduction, dual incremental and 19 bit absolute encoding, and in the shoulder case a holding brake, packaged so that ten fit inside a 66 kg vehicle [2].