OSAM-1

Michael Guinto. Public domain (NASA / US government work).

Overview

OSAM-1 was a NASA technology demonstration mission intended to rendezvous with, refuel and relocate Landsat 7, and to demonstrate assembly of a communications antenna with a robotic arm [8]. It began in 2015 as Restore-L, the Robotic Servicing Demonstration Mission, within the Technology Demonstration Missions program of the Space Technology Mission Directorate, and was renamed OSAM-1 in April 2020 when the SPIDER payload was added. NASA discontinued it in March 2024 following an independent review board recommendation [8].

The mission was organized as a servicing spacecraft carrying two separate robotic payloads. The project was assigned to NASA GSFC; the spacecraft bus was procured from Maxar; SPIDER carried Maxar’s Dragonfly antenna assembly experiment, added in 2019 and 2020 under a NASA Tipping Point award; the client, Landsat 7, is operated by USGS [8]. The FY 2024 Consolidated Appropriations Act directed $227M toward assessing a descoped, 2026-launch version of the mission [31].

Specifications

Parameter	Value	Source
Robotic arms	3 total: 2 on the Servicing Payload, 1 on SPIDER	[6]
Servicing arm reach	2.46 m, 7 DOF, six-axis force/torque sensor	[7]
SPIDER arm	5 m, 7 DOF, MDA-provided dexterous end effector	[7]

Mission profile

Parameter	Value	Source
Mission class	Category 1, Class C technology demonstration	[8]
Client	Landsat 7, non-cooperative, not designed for servicing	[7]
Propellant to be transferred	approx. 30 gal (114 l) of hydrazine	[1]
Antenna to be assembled	3 m Ka-band reflector from 7 elements	[6]
Beam to be manufactured	10 m carbon fiber composite	[7]
Started	2015, as Restore-L	[8]
Discontinued	March 2024, on an independent review board recommendation	[8]
SIR	November 2023, at NASA GSFC	[8]
Cost at cancellation	$1400M spent, $2380M estimate at completion	[8]

Design space and flight heritage

A servicer has to close from tens of kilometers to contact with a satellite, make a mechanical connection between two free-flying bodies, and then do work through that connection, here moving hydrazine. Microgravity removes the support that ground manipulation relies on: every contact force moves the servicer, and a relative navigation fault cannot be recovered by stopping. The corpus holds work on each stage, and the flown record is narrower than the research record.

Rendezvous and docking has flown. ETS-VII made the first unmanned to unmanned autonomous docking in three experiment flights between July 1998 and October 1999, with 7 mm lateral error against a 25 mm requirement and a peak contact acceleration of 14 mm/s2; a recurring thruster anomaly was handled by rewriting onboard software [11]. DARPA framed Orbital Express as a standard for satellite-to-satellite servicing interfaces rather than a single demonstration [16]. Its 2007 flight tests used a servicer and client flown as a pair, with stationkeeping from 4 km to 10 cm, five day and night captures at solar beta angles from 3 to 48 degrees, three direct captures, two grapple and berth operations and a commanded abort at 10 m [13]. The three-arm capture mechanism was qualified with hardware-in-the-loop contact tests and more than 400 simulated capture cases [14]. The video guidance sensor held range standard deviation below 2.5 mm against a 12 mm specification, though ground testing stopped at 100 m against a 200 m requirement [15]. OSAM-1’s client was not designed for servicing [7], which is the step beyond this record.

Arm operation on orbit has flown in limited form. ETS-VII telemetry showed generalized Jacobian control holding hand orientation in the inertial frame while the base pitch rotated 0.6 deg from arm reaction, and that free drift at 550 km exceeds 1 deg within a few minutes, which bounds how long attitude control can be released during a capture [12]. DLR’s ROKVISS, a two-joint torque-controlled manipulator, has operated on the ISS Russian segment since February 2005, and the same group framed the TECSAS free-flyer capture mission as the next step [17]. A planned ISS payload would have exposed rendezvous sensors to real lighting and targets using the station arm [28].

Capture of a tumbling or non-cooperative target has been shown on the ground only. A Canadian Space Agency laboratory demonstration captured a free-flying tumbling mock satellite autonomously with a two-arm testbed [18], and a dual-arm method captures a spinning object by repeated impacts, which tolerates unknown target mass and inertia [25]. Momentum-dumping control after accidental contact [24] and its center-of-mass extension [30] address the base reaction problem with thrusters used only to extract accumulated momentum. None of these has a flight record in the corpus.

Commercial concepts split by interface. SMART-OLEV was designed to dock through the client’s apogee engine nozzle and take over attitude and orbit control [22], and an 18-month DLR study argued that on-orbit failure rates make robotic servicing of geostationary satellites commercially essential [21]. NASA’s capability catalog groups this work into robotic manipulation, rendezvous and capture, relocation, repair, refueling, and manufacturing and assembly [19]. The 2022 edition shares whole paragraphs with the 2021 and 2025 editions, so the three are one source rather than three [20].

Mobility

OSAM-1 had no surface mobility. The spacecraft held six degrees of freedom of vehicle control [1] and was to reach the Landsat 7 orbit through a phasing maneuver, with a nominal 32 day autonomous rendezvous and docking phase [4]. Unlike a free-floating base that lets contact torques dump into vehicle motion, an actively attitude-controlled servicer such as OSAM-1 fights every capture and hose-deployment disturbance with its own thrusters; ground research into the alternative, letting a servicer’s center of mass and angular momentum drift after contact and regulating them back with a cascade controller, has been demonstrated only in hardware-in-the-loop rigs, not flight [30].

Manipulation was the mission’s motion problem. Three robotic arms were to fly: two on the Servicing Payload and one on SPIDER [6]. The NASA Servicing Arm has a reach of 2.46 m in seven degrees of freedom and terminates in a six-axis force/torque sensor [7]. Its design leans heavily on the flight-qualified arm developed for DARPA’s SUMO and FREND programs in the mid-2000s, carrying over motion control, robotic software frameworks, flexible harnesses, the force-torque sensor, joint design and flight operations experience. FREND itself was a 1.8 m seven-degree-of-freedom arm driven by the MOOG Rikishi Electronics Unit, which drives up to nine separate robot joints [7]. That heritage sits inside a longer lineage of on-orbit servicing manipulators, from the Shuttle and station arms through Orbital Express, that a 2014 survey of the field catalogs by mission phase: target observation and approach, final approach, capture and post-capture stabilization [9]. SPIDER carried a lightweight 5 m arm of seven degrees of freedom with an MDA-provided dexterous end effector, known as Dragonfly during its ground demonstration phase under a NASA Tipping Point partnership [6].

Power and energy

No figure for array output or battery capacity is published for the OSAM-1 spacecraft. In the flown configuration hydrazine and helium were supplied to the Servicing Payload propellant transfer subsystem across the spacecraft bus interface [2].

Thermal

Thermal design is published only at the level of the propellant path. The Hose Subassembly carries heaters and temperature sensors with EMI protection, a 28 VDC filter having been added after EMI testing in August 2017, and a thermal coating sized for full sun exposure with hydrazine inside the hose [1]. The Hose Management Assembly was qualified in thermal vacuum over 12 cold and hot cycles at qualification levels, followed immediately by 50 abrasion cycles, and separately over an additional 150 ambient cycles for more than 200 total. A one meter reference hose was tested in thermal vacuum after an ascent vent test and showed no change to the required thermal performance [2]. The Propellant Transfer Subsystem was required to keep both servicer and client propellant tanks within temperature constraints.

The precursor payloads set the thermal approach. The Robotic Refueling Mission Phase 1 payload used a cold biased design with a radiator on the back of the enclosure, multilayer insulation, and heaters sized for worst case cold [26]. For RRM3 cryogen transfer, autogenous pressurization with a heated capillary wick was tested using shuttle tile as the wick, in which liquid nitrogen wicks about 2 cm in one g [27].

Compute and avionics

The avionics element is described in the technology portfolio as reconfigurable avionics and software built around the SpaceCube processor, with video distribution and storage [1]. The robot system comprised the NASA Servicing Arm, a Robot Electronics Unit and robot flight software. Pressure transducers in the Clamshell Subassembly hold a combined component and avionics measurement error below 3 percent [1].

Flight software was one of the elements the independent review board named as carrying substantial remaining technical risk, alongside the Kodiak lidar and the Servicing Payload build [8].

Autonomy

Rendezvous and proximity operations used a relative navigation sensor suite spanning visible, infrared and lidar [1]. The client made the problem harder than a cooperative one: Landsat 7 has no fiducials for proximity operations, no grapple feature, and its fluid fill valves are closed out, each of which added complexity to the mission [7].

The division of autonomy was fixed by mission phase. Rendezvous, inspection and autonomous capture of a spacecraft not designed for servicing were Level 1 requirements met on board [8]; servicing was teleoperated from the ground [5]. Refuelling was to run from an onboard transfer script that calculates mass flow rate and totalizes transferred mass, stopping flow when a transfer duration, a totalized mass or a maximum pressure limit is reached, with a backup plan of discrete ground commands for conservative time durations and mass estimated on the ground while flow is stopped [1].

Hose deployment was a cooperative motion problem between the arm and the Propellant Transfer Subsystem. The arm and the roller-drive motor never operate simultaneously; motion alternates between the hose subassembly and the arm through predetermined trajectory waypoints, and no precision accuracy is required at a waypoint stop except for confirming hose deployment [1]. Deployed length is confirmed visually from ink marks on the conduit at 720 degree intervals, backed by position telemetry from the hose motor and by a situational awareness camera on the vision subsystem [1], [2].

Communications

Near-continuous Ka-band access was a mission requirement, and TDRS availability in the launch timeframe was flagged as a potential problem because only TDRS 8, 10, 11, 12 and 13 remained and their operational state was uncertain [8].

The relative navigation sensor suite traces its scope to an earlier servicing anomaly. On Orbital Express in May 2007 a corrupted relative navigation filter produced an unplanned separation of up to about 6 km between the servicer and client, and operators had to fall back on ground radar at Haystack Observatory to relocate the two vehicles; DARPA’s recommendation afterward for near 4 pi steradian relative sensing is why OSAM-1 added the Kodiak lidar [8]. By November 2023 Kodiak was itself about 11 months behind schedule, with a commercial backup lidar that would not arrive until after nearly all spacecraft environmental testing was complete [8].

Payload and instruments

The Servicing Payload carried two robotic systems, each arm terminating in an Advanced Tool Drive System with three rotary torque interfaces, electrical interfaces, actuators and two cameras for precision views [7]. Tools included a gripper, blanket cutter, wire cutter, cap removal tool and nozzle tool [1].

The Propellant Transfer Subsystem was split into six subassemblies across two assemblies [1]. The Propellant Transfer Assembly comprises the Hydrazine Transfer Subassembly and the Vent Thruster Subassembly; the Hose Management Assembly comprises the Roller-Drive, Hose, Clamshell and Box subassemblies.

PTS element	Function and specification	Source
Hydrazine Transfer Subassembly	isolates bus hydrazine, measures flow rate and totalized mass, controls flow; maximum flow set by a cavitating venturi	[1]
Flow meter	one of two unique developments for OSAM-1, alongside the flexible hose; flight units showed the required flow rate accuracy across the required flow rate and viscosity ranges before and after environmental testing	[2]
Vent Thruster Subassembly	GPM flight-spare monopropellant thruster, series-redundant single-seat solenoid valves, vents GHe and N2H4/GHe mixtures	[1]
Hose Subassembly	approx. 4 m, annealed stainless convoluted inner hose, 100 percent radiographed, 5x operating pressure burst rating	[1]
Hose deployment rate	0.2 or 0.5 cm/s via the roller drive	[1]
Hose qualification	verified to hundreds of allowable mechanical twist, torque and bend radius cycles; qualification and cleanliness certification completed June 2019	[2]
Box Subassembly	deflects about 3.8 mm (0.15 in) under 3-sigma launch loads; vent holes control ascent depressurization	[1]
Leak detection	1e-2 sccs GHe, an order of magnitude better than visual detection	[1]
Hypergol Refueling Tool	quick disconnect mating to a client fill and drain valve, hydrazine, extensible to other storable hypergols	[2]

The published hose deployment rate changed between accounts: the 2020 description gives 0.2 or 0.5 cm/s [1], the FY 2024 description about 0.35 cm/s in flight and 0.5 cm/s on the ground [2].

SPIDER was to assemble seven elements into a functional 3 m communications antenna and demonstrate Ka-band transmission to a ground station [6]. MakerSat, a Tethers Unlimited payload, was to manufacture a 10 m carbon fiber composite beam on board; it was descoped from the mission in 2023 [7].

Modes of operation

Operations center staffing was planned against mission phases of set duration [4].

Phase	Nominal duration
Launch, transit and checkout	30 days
Autonomous rendezvous and docking	32 days
Servicing	30 days
Departure	14 days
Assembly and manufacturing	85 days
Decommissioning	not stated

Source: [4].

Refuelling itself was sequenced against adiabatic compression risk. Initial Landsat 7 priming runs with the OSAM-1 hydrazine tanks at launch pressure, which is below nominal transfer pressure, and the servicer tanks are pressurized to nominal transfer pressure only after the manifold is wetted [1]. Before disconnect the transfer manifolds are vented and evacuated through the vent thruster to minimize the volume of hydrazine released.

Ground operations

Servicing was teleoperated from the ground under a telemetry delay of several seconds, which the interface studies emulated with a fixed 5 second round-trip delay [5]. Operations were to run from a single large mission operations center of more than 50 console positions, with a backup center available to offload peak staffing [4].

Human factors analysis of the operations center set the console layout and the shift pattern. Shifts were planned at 12 hours on console with 12 hours off between them as the figure that allows commute and rest, with fatigue and sleep monitoring and a schedule that lets biological rhythms adapt [4]. Mission-time clocks were sized at about 16 feet wide for legibility across the room, control room lighting followed ISO 11064-6:2005, and keyboard height, distance and display position were required to be adjustable for operators doing extensive data entry.

Interface research targeted the latency problem directly. A virtual reality planning environment splits the task into an interactive planning phase, in which the operator builds a motion plan and previews it in a virtual environment, and a supervised execution phase, in which the plan runs and is watched in augmented virtuality with an early stop available [5]. In a nine-operator study against a baseline keyboard and mouse two-dimensional interface, a commercial head mounted display gave a lower NASA TLX workload, 2.39 against 3.21 with a p value of 0.070, and lower reported difficulty, 1.56 against 2.56 on a five-point scale, with no loss of performance: all nine operators completed the task under both interfaces. The earlier three-dimensional interface built on a da Vinci surgical console had worse outcomes and higher workload than that same baseline.

Ground test infrastructure was extensive. The Goddard Servicing Testbed carries a Landsat 7 mockup and was rebuilt around flight-like engineering test units of the Hose Management Assembly, latch valve, flow meter and clamshell subassembly, with software mimicking the flight controls; it returned to service in late 2023 for integrated concept of operations testing and operator training [2]. The Kennedy propellant transfer testbed is an end-to-end functional and volumetric simulator of the flight subsystem used for flow, pressure surge and evacuation and vent testing. Integration and test was to be conducted across an AutoCapture Testbed, a FlatSat Testbed and lidar vacuum testing at Marshall [8].

Hardware-in-the-loop rendezvous facilities fill the same role for proximity operations. DLR’s EPOS 2.0 reproduces the six degree of freedom relative motion of two spacecraft with two industrial robots, one on a 25 m rail, at 1.56 mm and 0.2 degrees accuracy at 3 sigma [23], with an admittance law that matches the robots’ impedance to the simulated satellites during contact [29]. The free-floating control work cited above used DLR’s OOS-Sim, in which a model drives an industrial arm, so the satellite dynamics match the controller’s assumptions exactly [30].

The Robotic Refueling Mission on the ISS was the operational precursor, running from 2011 to 2021 across three phases and using Canadarm2 and Dextre to work the tools [3]. RRM1 was a 550 lb payload of 33 by 43 by 55 in launched on STS-135 that transferred 1.7 liters of ethanol; RRM2 added five adapters actuated by the multi-function tool and demonstrated cutting lock wires, removing valve caps, mating electrical connectors and manipulating thermal blankets; RRM3 was a 700 lb module of 30 by 45 by 45 in launched on CRS-16 in December 2018 for cryogenic transfer. Each RRM tool carried two orthogonal cameras and ARToolKit fiducials with calibration patterns.

Technologies developed

The RRM sequence produced the tool designs that became the OSAM-1 adapters, at lower mass and storage volume, and matured the machine vision and fiducial approach used for on-orbit alignment [3]. Its operational lessons were explicit: high-fidelity ground hardware is what makes on-orbit operations succeed, ground troubleshooting keeps an operation from being terminated, machine vision fiducials should be metrologized as accurately as possible, scripted command groups reduce errors, procedures should carry expected task durations, and planned breakout points stop a delayed task from cascading through the timeline.

The Hypergol Refueling Tool is licensable through NASA’s Technology Transfer Program [7]. The turbine flow meter, flexible servicing hose, hypergol refueling tool and cooperative servicing valve are all listed as OSAM-1 technical developments feeding the wider in-space fluid transfer portfolio [2]. The Advanced Tool Drive System, the Hose Management Assembly and the Hypergol Refueling Tool are cataloged as concluded ISAM hardware developments rather than as lost work [7].

Cancellation

The independent review board reported on 29 February 2024 [8]. Its cost analysis put the estimate at completion at $2380M against approximately $1400M already spent by November 2023, implying $980M to go, and put the 70 percent confidence launch readiness date at March 2028. Projected cost had risen from $1780M in 2020 with a September 2025 launch readiness to $2280M in November 2023 with July 2027 readiness, with no margin retained against that estimate [8]. Goddard head count was at or near 600 full-time equivalents in November 2023, 80 percent contract support, against a burn rate of $13M to $14M per month. Against an original forecast of no more than $750M and a 2020 launch, the project was seven to eight years late and $1500M over budget [8].

The board’s substantive finding was about the mission rather than the execution. Stakeholders across civil, defense and private sectors expect future national security spacecraft to be refuelled through prepared interfaces, so the industry had evolved past OSAM-1’s primary objective of refuelling a legacy vehicle through its fill and drain valves [8]. The board recorded that dexterous robotic servicing will be an important national capability and that completing OSAM-1 would have demonstrated key aspects of it, but recommended discontinuation on the grounds that the specific goal of delivering propellant to an aged Landsat 7 was not perceived as valuable against the cost to go.

MRV, developed through DARPA’s RSGS partnership with Northrop Grumman SpaceLogistics, uses the same FREND arm heritage and was to begin servicing unprepared GEO clients from 2026 [7]. Northrop Grumman’s own account names the RSGS arm the Flight Robotic Arm System 1, a seven-degree-of-freedom design from the Naval Research Laboratory exercised through its full range of motion in a ground trial called the Gauntlet before MRV’s 2026 launch [10].

What is not established

No hydrazine was ever flowed through the flight propellant transfer system and no on-orbit data exists; every flow, pressure drop and leak figure is a water or helium measurement scaled to hydrazine, and the underlying curves are not published [2]. The 2020 and FY 2024 accounts disagree on hose rate and leak detection, so neither is a final flight value [1], [2]. The review board’s report is a programmatic assessment with no subsystem test results, and its market judgments rest on a limited set of interviews [8].

Operator-interface evidence is thin. The virtual reality study had nine participants and one task, no difference reached p=0.05, and no communication latency was imposed [5]. The operations lessons are qualitative and camera specific [3], and the human factors report was advisory, with no workload data [4].

Servicing of a non-cooperative, non-prepared client by a free-flying arm is not demonstrated in flight anywhere in the cited record. The free-floating control results come from ground emulation with single runs and no thruster model [30].

References