Orbital Express

DARPA. Public domain (US government work).

Overview

Orbital Express was a DARPA demonstration mission flown in 2007 [11] to establish that satellite servicing could be performed by machines with no crew present and, for most of the mission, no human in the control loop. ASTRO, the servicer, was built by Boeing Phantom Works, which also supplied the end-to-end guidance, navigation and control system [1], [2]. NEXTSat, the client, was built by Ball Aerospace on its commercial RS-300 bus and played two roles at once: a satellite in need of servicing, and a commodities depot from which ASTRO could draw. The program was run by the DARPA Tactical Technology Office to validate the technical feasibility of robotic autonomous on-orbit refuelling and reconfiguration of satellites [11], a goal DARPA still cites the mission for [12].

Specifications

Parameter	Value	Source
Arm	Orbital Express Demonstration Manipulator System, 6 DOF, 2.8 m reach, MDA	[6]
Arm design family	scalable 3 m to 6 m, maximum reach about 3.3 m in the 2003 description	[1]
Arm end effector	mouse-trap cone and probe, with an end effector camera	[6], [1]
Docking mechanism	three-arm grapple, motor-driven lead screw, Starsys Research
Docking axial capture distance	15 cm
Docking angular misalignment tolerance	+/-5 deg in pitch, yaw and roll
Docking lateral misalignment tolerance	+/-5 cm
Docking contact velocity tolerance	3 cm/s
Mated preload	11,250 N
Capture time	under 10 s; capture and latch under 240 s
Docking interface outer diameter	under 46 cm
Docking mechanism mass	under 23 kg active, under 11.5 kg passive

Rows with no marker are from [5].

Mission profile

Parameter	Value	Source
Launch	Atlas V from Cape Canaveral SLC-41, 8 March 2007, both spacecraft mated	[2]
Orbit	492 by 492 km at 46 degrees inclination	[2]
Decommissioned	22 July 2007	[2]
Captures flown	5	[2]
Mate and demate cycles flown	9	[6]
Orbital replacement units transferred	2, a lithium-ion battery and a flight computer, both functioning flight hardware	[1], [6]
ASTRO hydrazine loaded	72 kg, of which 37 kg was designed for transfer	[1]
NEXTSat fluid transfer module tank	34 kg capacity	[1]
Hydrazine transfer flow rate error	5 to 10 percent, no anomalies	[6]

Subsystem responsibility was distributed. Northrop Grumman Space Technology supplied the fluid transfer and propulsion subsystem, MacDonald Dettwiler and Associates the robotic arm, Starsys Research the soft capture mechanism, Draper Laboratory the onboard mission manager, and NASA Marshall the Advanced Video Guidance Sensor [2].

The prior orbital servicing demonstration with a manipulator, ETS-VII, operated a 2 m arm against a target it had released itself, under ground supervision [6]. Orbital Express performed autonomous rendezvous from hundreds of kilometers, autonomous capture of an unconstrained free flyer, autonomous propellant transfer, and autonomous exchange of functioning avionics between two spacecraft [6].

Kinematics and workspace

The Orbital Express Demonstration Manipulator System was a six degree of freedom arm with 2.8 m reach, developed by MDA [6]. The 2003 program description gives a maximum reach of approximately 3.3 m and states the design was intended to be scalable from 3 m to 6 m, with ASTRO eventually carrying an interchangeable toolkit of end effectors [1], so the flown arm was at the short end of the design family.

Joint design and actuation

The arm avionics are identified as a Manipulator Control Unit and the ground element as the Manipulator Ground Segment. No joint, gearbox or motor description is published.

End effector and capture interfaces

Orbital Express used two separate mating interfaces, one for arm-mediated capture and one for vehicle-to-vehicle docking.

The robotic arm end effector grappled a fixture on NEXTSat and grapple fixtures on the two orbital replacement units. The capture interface is characterized as a mouse-trap, cone and probe arrangement [6]: a probe entering a cone that provides passive lateral and angular correction, latched by a sprung mechanism. The arm also carried a camera on the end effector [1].

The docking mechanism on ASTRO, built by Starsys Research and now marketed by Sierra Space, was a three-arm grapple that made the structural, electrical and fluid connection once the two vehicles were in contact. An active half on ASTRO carries a motor-driven lead screw actuating three linkages; a passive half on NEXTSat presents the features those linkages grab [5]. Closing the linkages draws the passive structure into a three-point kinematic mount, aligning cup and cone features, after which the motor continues to apply preload until the interface reaches the required stiffness. Release reverses the sequence and the spring-loaded kinematic mounts supply separation velocity.

Preload is set by a torque-sensing mechanism: the motor runs until a mechanical torque sensor trips limit switches at the torque corresponding to the required preload, at which point the motor stops and its brake holds the load. A sensor on the passive side indicates proper engagement of the grappling arms. Fluid and electrical couplings, including floating electrical connectors and a self-sealing quick-disconnect fluid coupling, sit inside the interface ring so that mating the structure mates the utilities [1], [5].

The mission demonstrated nine mate and demate cycles on orbit [6]. The ground program behind that number was substantial: more than 200 mechanical mate and demate cycles during prototype development, over 400 simulated capture cases in a dynamics model covering a uniform distribution of initial conditions, hardware-in-the-loop six degree of freedom testing on a hydraulic hexapod with a force and moment sensor in the load path, and micro-gravity aircraft flights that exercised grapple and capture within the 25 second zero-g window [5].

Force and moment sensing

No arm force or moment sensor is published. Docking loads were managed mechanically rather than by force feedback: the soft-docking approach was chosen over impact docking and over a harpoon or reeled-probe configuration precisely so that contact forces stayed low enough to align fluid and electrical couplings safely, and the preload is closed on a mechanical torque sensor rather than a load cell [5].

Reach, payload and what was transferred

Two orbital replacement units were carried, both functioning flight hardware rather than mass models: a lithium-ion battery and a flight computer [1], [6]. The arm transferred the battery back and forth between the two spacecraft and removed and replaced the operational computer, and these operations were demonstrated individually and in combination without ground assistance [2].

Propellant transfer used hydrazine. ASTRO carried 72 kg of monopropellant hydrazine and was designed to transfer 37 kg; NEXTSat carried a modular fluid transfer module with a 34 kg tank capacity [1]. Transfers were run in both directions so that each vehicle acted in turn as supplier and recipient. Multiple refuelling demonstrations were performed before the first separation and after each of the five captures [2]. Flow sensors showed 5 to 10 percent flow rate error on the hydrazine transfers, with no anomalies reported [6].

Power and energy

ASTRO used deployable gimballed solar arrays and two 33 A h lithium-ion batteries [1]. NEXTSat power is not described beyond its RS-300 bus heritage.

Thermal

ASTRO’s structure was aluminum honeycomb panel construction over a central cylinder backbone with built-in thermal radiators, sized for the mated ASTRO and NEXTSat launch configuration [1]. No operating or survival temperature ranges are published. The orbit itself sets the cycling load: at 492 km an eclipse is entered roughly every 90 minutes, the cadence that produced about 34,000 cycles in 5.8 years on LDEF [8].

Compute and avionics

ASTRO carried a radiation-hardened PowerPC 750 processor in an open, modular architecture, with an IEEE-1394 plug-and-play embedded network for subsystem interconnection and distributed telemetry data acquisition [1]. A sun-safe mode was provided. The robotic arm had its own Manipulator Control Unit. Radiation hardening at 492 km and 46 degrees is driven by single event effects from trapped protons rather than by total dose, which is the same constraint that governs ISS avionics part selection [7].

The flight computer was itself one of the ORUs, which is a deliberate architectural point: the mission was intended to show that the most integration-sensitive box on a spacecraft could be swapped in orbit [1], [2].

Autonomy

The program defined four levels of supervised autonomy and flew operations across the range [1]:

1. Ground approval or data uplink required before execution.
2. Ample time allowed for ground override before the system executes automatically.
3. Autonomous execution, with commands sent to the ground for occasional verification.
4. Fully automated, with ground analysis only when a problem occurs.

Draper Laboratory supplied the onboard mission manager software [2]. The division of labor it embodied was that the ground kept overall mission planning and the onboard system executed the resulting task list, monitored its own performance and safed the vehicle when it detected an anomaly. An onboard-commanded abort during approach, at 10 m range, was one of the flight exercises [2]. Unmated operations ran with the ground in monitor and override mode only, and then only during communication passes.

ASTRO carried the Autonomous Rendezvous and Capture Sensor System: three visible cameras, one infrared camera, a long-range lidar, and the Advanced Video Guidance Sensor [1].

AVGS is a laser sensor that recovers full six degree of freedom relative pose from retroreflective targets. Two sets of laser diodes at 800 nm and 850 nm fire in turn through a mirror, and a camera images the returns [3]. Targets carry filters passing one wavelength and blocking the other, so subtracting the second image from the first and applying an intensity threshold removes essentially all background clutter, including sunlit structure [3]. Retroreflectors are arranged in a pattern the software knows, which is what converts a set of bright spots into a pose. Separate long-range and short-range targets were flown, each keyed to one wavelength, with a nominal overlap region of 10 m to 30 m where both could be tracked [3]. Output is at 5 Hz with internal tracking at 10 Hz, the field of view is plus or minus 8 degrees, and the target must stay within a seven degree cone about the field-of-view center [3]. Five modes were implemented: standby, acquisition, track, reset and maintenance.

Boeing’s Vision-based Software for Track, Attitude and Ranging computed azimuth, elevation and range to the client from the camera imagery, and roll, pitch and yaw where the client presented sufficient natural features. Unlike AVGS, it required no cooperative markings on the target, which is the capability line that separates servicing a satellite built to be serviced from servicing one that was not [2].

Onboard fault protection covered the mated approach: the system could handle a problem arising during final mating operations, and an onboard-commanded abort was deliberately exercised in flight during an approach at 10 m range [2].

Communications

ASTRO used Air Force Satellite Control Network SGLS and TDRSS S-band links for command and control, with encryption and decryption, plus a dedicated crosslink subsystem to NEXTSat, which carried its own crosslink antenna [1]. Both ground-network and orbiting-relay communication access were exercised during rendezvous and capture, deliberately, since a servicer must be able to work through either [2].

Payload and instruments

Orbital Express carried no science payload. Its instruments are the rendezvous sensors described under autonomy, the arm end-effector camera [1], and the passive aids on NEXTSat: retroreflectors on the docking face usable from 60 m and closer, a passive docking sensor, and a grapple fitting for the alternative arm-based mating path.

Modes of operation

Two distinct capture methods were flown [1]:

• Direct capture, in which ASTRO flies into contact with the capture mechanism on NEXTSat and the mechanism engages.
• Grapple and berth, in which ASTRO holds a larger standoff, reaches out with the robotic arm to grapple the client, and draws it in until the capture mechanism can engage. This is the mode that demonstrates capture of an unconstrained free flyer.

Three direct captures and two grapple-and-berth captures were performed [2]. Grapple and berth places the load path through the arm and the grapple fixture; direct capture places it through the docking mechanism, whose 11,250 N mated preload and 3 cm/s contact velocity tolerance were sized for that role [5].

The rendezvous and capture campaign was designed as a matrix of conditions rather than a repetition of one profile [2]. Approaches were flown from behind and in front of the client, from distances up to 410 km, on stable-orbit and co-elliptic trajectories. Two stationkeeping algorithms were demonstrated on various sides of the client at 4 km, 1 km, 500 m, 120 m, 30 m, 10 m and 10 cm; inspection flyarounds were flown both elliptical, plus or minus 120 by 60 m, and near-circular at 100 m, at one and three times orbital rate; and captures were performed in solar-inertial, plus-V-bar and minus-R-bar attitudes, two in daylight and three at night, across solar beta angles from 3 to 48 degrees [2].

The flight plan was written as a flexible template. Issues encountered during the second rendezvous and capture exercise caused that exercise to expand unintentionally and to pick up objectives assigned to later exercises, which let the planned seven exercises be reduced to five. A sixth exercise was then appended at the end of the flight, deliberately producing separated orbits whose trajectories could never cross again [2].

Ground operations

Command and control ran from the Air Force Research Laboratory Research, Development, Test and Evaluation Support Center at Kirtland Air Force Base, Albuquerque. An engineering support room operated at Boeing Huntington Beach, California, and a rendezvous support room at Boeing Tower 2 in Houston [2]. The manipulator had its own ground element, the Manipulator Ground Segment [1].

Because the vehicles executed autonomously between passes, the ground role during unmated operations was monitoring and override rather than commanding [2]. Autonomy level, not link capacity, set that division: at these ranges the round-trip delay is small, but contact between two free-flying vehicles happens faster than a ground loop can respond, and the onboard mission manager held authority through the approach.

The daily and long-range mission plans behind that authority were generated, not written by hand: JPL’s ASPEN scheduling tool produced the timeline each vehicle executed, working against a domain model of Air Force Satellite Control Network and TDRSS ground-station contact rules, including a 60-minute reuse delay before a ground string already used could be re-tasked and a re-planning cadence that ran from 24 to 7 days out down to a same-day pass [9]. That is the layer that made the four supervised-autonomy levels workable in practice: the plan itself, not just the vehicle’s onboard logic, had to absorb schedule slip and lost contacts without ground intervention.

The crewed alternative shows what that bought. Space Shuttle rendezvous and proximity operations, flown for three decades against a comparably wide range of target vehicles, stayed crew-flown in translation and hand-controlled at close range; the flight dynamics organization that operated it attributes this to the absence of a programmatic requirement for automation and the cost of certifying safety-critical flight software, not to a sensor limitation, and argues that continuous out-the-window observation of the target was needed to catch off-nominal relative motion [10]. Orbital Express was built specifically to answer that requirement, on an uncrewed vehicle where a certified flight-critical human interface was not an option to begin with.

Technologies developed

The mission retired specific risks rather than producing general-purpose hardware. Its outputs were autonomous rendezvous and capture, component change-out, refuelling and orbital control demonstrated together on one flight, which is what made servicing credible enough for follow-on programs to be funded [2].

The mechanism’s design lineage runs from a 1999 AFRL Phase I SBIR for autonomous soft docking, through a Phase II SBIR, to the Orbital Express flight unit [5].

AVGS went on from Orbital Express, having tracked its short-range target continuously to 31.6 m during 30 m unmated operations and acquired its long-range target at 150 m during a recovery approach, well past the planned 120 m limit [3]. The AVGS lessons-learned are unusually direct about why it did not do better: ground optical testing was capped at 100 m against a 200 m requirement, parameters beyond 100 m were extrapolated, and the shortfall past 150 m is attributed squarely to the missing ground tests rather than to the sensor [4]. Where ground testing had been done, ground and flight data agreed well.

Boeing carried the guidance, navigation and control system forward into a second-generation autonomous rendezvous and capture system, updating sensors, improving Vis-STAR target modeling and navigation algorithms, and upgrading the sensor and avionics laboratories in Huntington Beach and Houston [2]. NASA Goddard formed the Space Servicing Capabilities Project in 2009 explicitly to build on the Orbital Express experience, leading to the Robotic Refueling Mission payloads on the ISS and eventually to OSAM-1.

Autonomous capture of a non-cooperative vehicle is identified in the mission reporting as the remaining challenge. NEXTSat carried retroreflectors, grapple fixtures, a matched docking interface, a crosslink radio and commandable control modes; relative navigation against targets with none of those was not demonstrated at all ranges [2].