Robotic Refueling Mission

NASA/David Saint-Jacques. Public domain (NASA / US government work).

Overview

Servicing a spacecraft that was never designed to be serviced means working through interfaces built for a technician’s hands: lock wire, thermal blanket closeout tape, valve caps sealed for a launch and never a reopening. The Robotic Refueling Mission was NASA and the Canadian Space Agency’s multi-phase effort to identify and develop the tools, techniques and procedures for refuelling and servicing such a satellite from a robotic manipulator instead [1]. It combined the robotic systems of the International Space Station with the tool design and satellite servicing work of the Exploration and In-space Services division at Goddard Space Flight Center. Between 2011 and 2021 two external station payloads housed more than a dozen robotic tools and adapters [1]. NASA’s public overview page for Phases 1 and 2 describes the RRM module as a washing-machine-sized cube of about 550 pounds and 33 by 43 by 55 inches, carrying activity boards and four stowed tools plus 1.7 liters of ethanol for the fluid transfer demonstration [17].

The tasks demonstrated were cutting lock wires, manipulating fasteners and thermal blanketing, removing caps, connecting to and resealing a fuel port, transferring fluid in zero gravity, and long-term storage of cryogens [1]. Phase 1 added the operation of typical door latches and the mechanical manipulation of valves on a dummy coolant valve panel task board [2]. Phase 1 addressed the steps needed to refuel a legacy spacecraft with storable propellants and the early steps for replenishing cryogens; Phase 2 addressed intermediate steps for cryogen transfer; RRM3 addressed the final steps.

Every task was executed by robot operators at NASA JSC and at CSA using the Special Purpose Dexterous Manipulator, which retrieved and operated the tools from the payload [1]. SPDM is itself one of the station’s two large manipulators, alongside the Space Station Remote Manipulator System it rides on; survey comparisons of flown space manipulators place Canadarm2 and the newer Chinese Space Station arms in the same lineage of large dexterous servicing arms that RRM’s tools were built to work with rather than against [18]. SPDM’s other flown role illustrates how central that platform is to what RRM could attempt: a proposed Dextre Pointing Package would have mounted rendezvous and docking sensors, including a triangulation and time-of-flight lidar unit, directly on the arm and used closed-loop arm control to track visiting vehicles, on the reasoning that a full rendezvous envelope spans kilometers of range and nearly 4 pi steradians of attitude, which only a real orbital platform, not a ground rig, can exercise at full scale [13]. The Phase 1 module reached its work site in three steps: an EVA crew member moved it from the Shuttle payload bay to the SPDM Enhanced ORU Temporary Platform, then the SSRMS used the SPDM to relocate it to ELC4 [2].

Specifications

Parameter	Value	Source
Phase 1 module mass	250 kg (550 lb)	[1]
Phase 1 module envelope	838 x 1092 x 1397 mm (33 x 43 x 55 in)	[1]
RRM3 Fluid Transfer Module mass	318 kg (700 lb)	[1]
RRM3 FTM envelope	762 x 1143 x 1143 mm (30 x 45 x 45 in)	[1]
Tool interface	SPDM ORU and Tool Changeout Mechanism	[2]
Cameras per tool	2, orthogonal	[2]
Common tool structure	Tool and Vision Support Structure, providing stowage, electrical and vision interfaces	[2]

Ethanol rather than a propellant was flown in Phase 1 to limit risk to the station and its crew [1]. RRM3 by contrast carried roughly 50 liters of liquid methane, a combustible cryogen, which drove both the ground servicing design and the launch site selection [4].

Mission profile

Event	Date	Source
Phase 1 module launched on STS-135, the final Shuttle flight	2011-07
Removed from the Shuttle payload bay by an EVA crew member, the last payload ever removed that way	2011-07
Installed on ELC4 by SPDM	2011-09
Phase 2 task boards and VIPIR launched on HTV-4 and ATV-5	2013-08, 2014-07
Phase 2 hardware installed on the RRM module by SPDM	2015-04
Phase 1 and 2 hardware removed from the station	2017-03
Phase 1 and 2 hardware destroyed on the re-entry of SpaceX CRS-10, except the SCT and Task board 4, returned as pressurized cargo	2017
RRM3 Fluid Transfer Module, three tools and a mounting pedestal launched on SpaceX CRS-16	2018-12-05	[1], [4]
SPDM installed the FTM on ELC1	2018-12
SPDM installed the tool pedestal and tools	2019-04
Cryocooler failure; stored liquid methane vented before robotic cryogen testing	2019-04
FTM scheduled for de-orbit	2022
Phase 1 fluid flown	1.7 liters of ethanol	[1]
RRM3 cryogen flown	approximately 50 liters of liquid methane	[4]

Rows with no marker are from [1].

Two EXPRESS Logistics Carriers hosted the payloads: ELC4 on the starboard-nadir side for Phases 1 and 2, ELC1 on the port-nadir side for RRM3 [1]. RRM3 launched on 5 December 2018 [4].

Tools and adapters

Item	Phase	Function
Wire Cutter Tool (WCT)	1	Cut lock wires between fill and drain valves; a small gripper grasps and manipulates thermal blanketing, and a blade cuts blanket closeout tape
Safety Cap Tool (SCT)	1	Remove and stow a fill and drain valve safety cap and its seal
EVR Nozzle Tool (ENT)	1	Open a fuel valve, transfer fluid through it, and close it
Multi-Function Tool (MFT)	1	Carrier for several small adapters, for tasks such as removing fuel valve caps or a gas plug
Electrical Plug Adapter (EPA)	2	Capture and remove a connector turnaround plug
Wire Harness Adapter (WHA)	2	Mate an electrical plug and run a connectivity check
Vent Plug Adapter (VPA)	2	Insert into a vent line to seal a gaseous line before refuelling
Coolant Line Adapter (CLA)	2	Mate a coolant fitting to an interface, tethered on a spring reel to simulate hose forces; no coolant transferred
Blind Mate Adapter (BMA)	2	Connect an array of SMA plugs in a receptacle box
VIPIR	2	Flexible tube with a camera at an articulatable tip, navigated through conduits of several materials to inspect inside a spacecraft
Cryogen Servicing Tool (CST)	3	Capture and feed a flexible, non-fixed cryogen hose into a receiver tank
VIPIR2	3	Upgraded cameras; the only RRM tool to transmit video over wifi rather than through the SPDM and SSRMS data lines
MFT2	3	First and only RRM tool with adjustable Camera Positioning Mechanisms, allowing the tool cameras to be extended or pitched per task
Cryogen Coupler Adapter (CCA), Xenon Coupler Adapter (XCA)	3	Fluid transfer through fixed dedicated supply hoses

Source: [1].

The MFT is the design argument the program makes about mass: exchangeable adapters rather than a separate tool for each task saves mass and volume on a servicing mission [1]. On Phase 1 the MFT carried four adapters, including a tertiary cap adapter for the refuelling task [2]. Exposing a fill and drain valve requires cutting away its closeout multi-layer insulation and removing the tertiary and safety caps with their safety wires cut, which is the sequence the WCT, SCT and MFT were built around.

Task boards carried the work sites [2]. Task board 2 held a tertiary cap and a safety cap with their wires for wire cutting and cap removal, both dummy caps not plumbed into the fluid transfer system, a test port panel for blanket manipulation, and three machine vision tasks: an explosive bolt hole simulation, a Hubble J-hook, and a Hubble door latch [2]. The boards also carried solar cell technology and materials coating demonstrations and machine vision fiducials, including the ARToolKit fiducial and a camera calibration pattern [1].

RRM3 cryogenic results

RRM3 demonstrated zero boil-off storage of about 19 kg of liquid methane for nearly six months, of which four months were on orbit [3]. The following were achieved or attempted.

Item	Result
Zero boil-off cryogen storage	Achieved, nearly 6 months total, 4 months on orbit
Radio Frequency Mass Gauge	Mass gauging achieved on the ground and in microgravity
Turbine flowmeter	Partially successful on the ground: accurate when it registered flow, but did not always register it; not used on orbit
Liquid and vapor detectors, of the type used on the Superfluid Helium On-Orbit Transfer experiment	Distinguished liquid from vapor on the ground and on orbit during storage; ambiguous during ground transfers because of fluid supersaturation
Autogenous pressurization by a wick and heater technique developed at GSFC	Achieved on the ground and on orbit
Integrated Multi-Layer Insulation	Allowed the receiver dewar cryocooler to reach below 82 K, against a methane triple point of 90.7 K
No-vent transfers with methane	5 achieved during ground testing at NASA KSC
Fluid management to orient liquid at the transfer line inlet in microgravity	Not tested, no on-orbit transfer occurred
Freezing a cryogen in microgravity	Not attempted on orbit; argon and methane were frozen in ground tests

The cryogenic portion of the mission ended early because of electrical problems [3], the proximate cause being an April 2019 payload power glitch that took out the cryocooler output circuits unrecoverably and forced the stored methane to be vented through the burst disk, blowing down to 5 psi in about 90 seconds [16]. Cryocooler behavior on the ground turned out to be a poor predictor of on-orbit behavior: cooldown and warmup time constants were much shorter, and the cold finger to bulk methane coupling weaker, than characterized at KSC, consistent with poor thermal strap contact to the bulk liquid in microgravity [3]. Development of the wick pressurization technique itself was done as a one-g surrogate: a Shuttle tile material wick lifted liquid nitrogen 1.25 cm above the liquid surface with up to 48 W applied, and a conduction model matched the measurements at high heater power but under-predicted heat into the liquid at low power, leaving the flight wick’s 200 liter per hour zero-g transfer target a design assumption rather than a tested rate [15]. Of the five ground no-vent transfer tests, three succeeded; the fifth liquid-locked the receiver hard enough to rupture an inner burst disk into the vacuum jacket, which the operations team failed to recognize for two days [16]. Most of the robotic tools were still maneuvered and operated as planned after the venting, giving results on the technique for transferring cryogens and manipulating the associated interfaces. Demonstration of the robotic tool interfaces was left to continue as the station schedule allowed, with no cryogen transfer [3].

Thermal

RRM uses a cold-biased design with a radiator on the back of the enclosure to hold temperature in the hottest environments, and multi-layer insulation plus heaters to hold component temperatures in the worst cold cases [2]. Heater power comes from three independent sources at three voltages, reflecting the three phases of the payload’s life. Analysis over the full ISS attitude envelope, including beta angles to plus or minus 75 degrees, closes only if operation is restricted: the design carries an ISS flight rule prohibiting RRM operation above 60 degrees beta, so part of the thermal margin is operational rather than hardware [2].

Source	Allocation	Notes
Shuttle, 28 V	230 W at maximum voltage, about 216 W at nominal	4 circuits on the fluid transfer system plate; each circuit controlled by two Elmwood mechanical thermostats in series, cycling between +25 and +30 degrees C
ISS, 120 V	290 W at maximum voltage, about 245 W at nominal	Two feeds, on the fluid transfer system plate, the fluid flex line, and the enclosure next to the plumbed fill and drain valves
OTCM, 15 V	5 W per camera, two per tool, and 10 W per tool electronics box	Camera thermostats set at -25 to -16 degrees C, tool electronics box at -32 to -26 degrees C

Values from [2]. Tool heaters are powered only through the OTCM, so no tool heater power is available while a tool is stowed on the tool stowage bench inside the RRM enclosure. The thermal cases are defined with the tools stowed and unheated for the non-operating cases, and with the MFT and ENT deployed and operating and the WCT and SCT stowed for the operating cases.

Ground operations and procedure development

Preparation for each phase ran for years before launch: goals were converted to preliminary hardware designs, prototypes of some tools and adapters were tested on industrial robots at Goddard to refine the design, and once high-fidelity CAD or hardware was available Goddard robot operators wrote draft procedures giving explicit detail on how to maneuver and operate the tools [1]. For RRM3 the ground campaign extended to building a methane servicing capability that did not previously exist [4].

Hardware testing was done in Goddard’s Servicing Technology Center, using a six degree-of-freedom industrial arm as a stand-in for one SPDM arm together with an OTCM Simulator supplied by MDA, the builder of the SPDM end effector [1]. Reviews of the wider on-orbit servicing field organize the capture problem this equipment supports into observation and planning, approach, impact and grasp, and post-capture stabilization phases, and note that free-floating manipulator dynamics of the kind SPDM exhibits when not base-controlled cannot be fully reproduced on the ground at all [10]. That combination let prototype, engineering design unit, flight spare and flight tools be operated as they would be in space, and supported both nominal and contingency procedure development. It is one instance of a broader pattern in on-orbit servicing programs: because true free-floating dynamics cannot be reproduced on the ground, every servicing developer substitutes a different approximation. DLR’s EPOS 2.0 facility uses two KUKA industrial robots, one riding a 25 m rail, driven by an admittance control law so their tip impedance during contact reproduces that of the simulated satellites, rather than Goddard’s approach of running flight-representative tools on a fixed industrial arm against a physical OTCM simulator [12]. Surveys of the field list a wider range of such rigs, from flat-floor air bearing tables to dedicated rendezvous and docking simulators, as the standard way the community closes the gap between ground validation and orbital dynamics [9] [10] [11] [14]. The tool thermal design assumes the same deployed-and-operating versus stowed distinction in its hot and cold cases [2].

Every RRM tool presents the same OTCM interface and the same Tool and Vision Support Structure, which is what makes one procedure format work across the set [2]. Goddard robot operators designed the robotic maneuvers and determined the OTCM commands, operators at JSC and CSA translated them for SSRMS and SPDM, and to make that translation straightforward every maneuver was expressed relative to the previous location rather than as Cartesian positions or joint angles [1] [17].

At least one campaign on MDA’s Ground Trainer was run for each phase, on robotic hardware and software representative of the station systems, against the same SPDM interfaces the tool thermal design assumes for power and stowage [2]; MDA engineers converted the Goddard procedures for the Ground Trainer robot first, and in some cases the torques, forces or step order in the Goddard procedures were not achievable with SPDM [1].

The Goddard robot control software includes a visualization of robot, tool drive, tools and payloads. For RRM3 it was extended to display and control simulated cameras, including the tool cameras, and to set the position and orientation of the tool tip and any attached adapter [1]. That allowed procedure development to start before hardware existed, and let tasks be evaluated with fewer people and less risk once it did; the simulation does not verify contact forces, OTCM turns and torques, or the motion of flexible hoses. Flexible hose handling is exactly what the Cryogen Servicing Tool was built to exercise, capturing a flexible, non-fixed cryogen hose and feeding it into a receiver tank.

Ground servicing of RRM3

NASA KSC loaded approximately 50 liters of liquid methane into the RRM3 payload’s source dewar before launch [4]. KSC had fueled rockets and payloads before but had never done so with liquid methane, so the ground support equipment and processes were developed from nothing: a 450 liter DOT-approved LNG microfueler supplied a purpose-built methane servicing panel, the payload sat on a high-precision scale for mass determination, and a nitrogen purge held payload oxygen below 1 percent whenever powered. Testing was moved out of the Payload Hazardous Servicing Facility into a smaller building to cut cost, with NFPA 497 Class 1 Division 2 zoning around the building and vent stack and a custom clean enclosure fed with ducted clean air [4]. The hazard profile is specific to a cryogen: the Restore-L and OSAM-1 propellant transfer subsystems that inherited RRM’s servicing approach instead handle storable hypergolic propellant, a different ground safety problem with its own seal-less pump and hose qualification work rather than LNG sourcing [16]. Requested grade B methane at 99.9 percent purity proved unobtainable in the United States at flight servicing time, and the payload flew with 98.2 percent no-grade LNG [4]. All ground operations were controlled locally by operators, with no automation and no software control of the ground support equipment. Pre-flight ground testing with liquid methane was performed both on an engineering demonstration unit and on the flight payload, also at KSC. Autogenous pressurization was demonstrated both on the ground and on orbit [3].

Technologies developed, and what came after

The program’s transferable outputs are the tool set and the operations method rather than a vehicle. Named results are the exchangeable-adapter architecture that replaces one tool per task [1] [2]; robotic cutting of lock wire and closeout blanket to expose a fill and drain valve that was never designed for servicing; a poseable borescope for interior inspection through spacecraft conduits [1]; wifi video off a robotic tool, avoiding the manipulator data path; adjustable camera positioning on a tool; and, on the fluid side, zero boil-off cryogen storage, radio frequency mass gauging in microgravity, autogenous pressurization by wick and heater, and integrated multi-layer insulation [3].

NASA states the objective as giving satellite owners the means to diagnose problems on orbit, fix anomalies and keep instruments working longer, and as seeding a commercial servicing industry [17]. RRM also established the station as a route to raise technology readiness for servicing hardware, and led to further technology partnerships between the Goddard servicing division and the ISS program [1]. NASA’s own capability catalogs place RRM inside a standing taxonomy of servicing functions, robotic manipulation, rendezvous and capture, relocation, repair and maintenance, and refueling and fluid transfer, that the agency has tracked since 2021 and carried forward with light revision under the renamed ISAM heading [7] [8] [18] [6]. The lessons are recorded as being applied directly to the propellant transfer subsystems of two follow-on missions. Restore-L’s Propellant Transfer Subsystem inherited a hose management assembly and a seal-less hypergolic pump aimed at the same problem RRM’s ENT addressed at small scale, refueling a satellite with no designed servicing interface, now for a full operational transfer rather than a demonstration [5]. OSAM-1 carried that architecture further before its cancellation: a qualified Hose Management Assembly deploying a metal hose at about 0.35 cm/s, a precision flow meter rated for microgravity and vacuum, and a Hypergol Refueling Tool sized to deliver about 30 gallons of hydrazine to Landsat 7, qualified through 12 TVAC cycles and 200 hose abrasion cycles [6]. On the cryogenic side the demonstrated items were autogenous pressurization, radio frequency mass gauging and integrated multi-layer insulation [3].

Lessons recorded

The published lessons are specific. Designing robotic tools for the space environment requires attention to materials, indicators, and the differences between ground and flight use cases and environments [1]; the thermal case is one instance, where a tool loses all heater power the moment it is stowed [2]. Mechanical status indicators are preferred over electronic ones, every action needs a confirmation especially where there is no recovery path, and tool finish and color must be chosen against the actual camera and backgrounds, down to a minimum indicator line of 0.030 inches and a dot of 0.060 inches diameter; machine vision fiducials fail under sunrise and sunset shadow conditions. The most expensive gap the program identified was between ground and flight robots: the rigid industrial arm at Goddard could not react to the large axial loads the Cryogen Coupler Adapter’s receptacle required, so flight-like practice for that specific task was never possible before flight [1]. Where the on-orbit operations window is limited, spending time and high-fidelity hardware on ground testing is what buys it back [1]; the RRM3 fluid campaign spent five ground no-vent transfers to that end [3]. Camera view quality, frame rate and position requirements are driven by the needs of troubleshooting rather than by nominal operations [1], which is the reasoning behind the adjustable camera positioning on MFT2 and the wifi video path on VIPIR2. Detailed hardware manuals, nominal and contingency procedures, and clearly defined operations team roles and protocols determine efficiency. The RRM3 ground servicing campaign reached the same conclusion from the fluid side, where every operation was run manually by operators with no software control [4].

What is not established

RRM3 never performed an on-orbit cryogen transfer: the April 2019 cryocooler failure forced the stored methane to be vented before fluid management at the transfer line inlet, no-vent transfer, or freezing a cryogen, could be attempted in microgravity, so none of those three objectives has a flight answer [3] [16]. The ground thermal design for Phase 1 rests on analysis against predicted limits, not on thermal balance or thermal vacuum test data or any on-orbit temperature measurement, and part of its margin is an operational restriction rather than demonstrated hardware margin [2]. The wick pressurization technique that flew was sized from one-g liquid nitrogen surrogate testing whose own authors say the underlying conduction model under-predicts heat into the liquid at low heater power, so the flight wick’s zero-g transfer rate target was a design assumption rather than a validated result [15]. The operations lessons on camera color, finish and indicator visibility are stated as camera-specific findings from this program’s operators, not as controlled comparisons, and are not established as transferable to other servicing camera systems [1].