European Robotic Arm

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

The problem

The Russian segment of the International Space Station needed a manipulator of its own. Canadarm2 and the Japanese Experiment Module arm cannot reach it, and base point and payload interfaces differ between segments [3]. The arm has to move payloads of up to 8000 kg, relocate itself along the segment, and do so at a mass low enough to launch, which makes it far more flexible than an industrial robot: the first bending mode holds over 90 percent of the total inertia and dominates the control problem [1].

The environment adds three constraints. Vacuum removes the liquid and air film that terrestrial brakes and bearings rely on, so friction pairs and lubricants must be qualified in vacuum [2], [5]. Orbital thermal cycling, about 34,000 eclipse transitions per 5.8 years in low Earth orbit as recorded on LDEF, produced microcracking in composites and delamination of coatings [6]. Meteoroids strike external hardware at an average 19 km/s [7]. A fourth constraint applies on the ground: the arm cannot support itself in one gravity, so three-dimensional operation cannot be tested end to end before launch [1].

What has flown

Shuttle Remote Manipulator System. The Shuttle arm flew 91 missions over 30 years, and its operability record includes uncommanded motion: during LDEF unberthing on STS-41C a pure Z command produced 2.5 inches of X motion over 11.5 inches of commanded Z, about 22 percent cross coupling [14]. A vacuum brake material problem is also recorded, in which an asbestos derivative lost torque through moisture loss [14].

Canadarm2 and Dextre. Canadarm2 has a dry mass of 1497 kg [8], and a survey lists it at 17 m with Dextre at 3.35 m and 7 DOF [34]. Dextre exchanges orbital replacement units in a three-step sequence, and the case for it is crew time saved on external maintenance; the 13.4 EVAs per year avoided come from a 2004 estimate and the paper predates commissioning [15]. Eleven years of Robotic Refueling Mission operations with Dextre-compatible tools cut lock wires, removed caps, mated connectors and transferred 1.7 liters of ethanol, but the lessons are qualitative, with no success rates or contact force measurements [36].

JEMRMS. The Japanese arm’s 1994 development paper tabulates structural eigenvalues: 0.32 and 0.34 Hz unloaded on orbit, 0.19 Hz with a 500 kg payload and 0.046 Hz with a 7000 kg payload, against 14.2 Hz restrained for launch [16]. These are design predictions made before the engineering model existed, to be updated by modal survey [16].

ERA. The arm launched on the Nauka module on 21 July 2021, originally planned for 2001 on the Space Shuttle [3], [13]. Its first planned tasks were airlock setup and radiator installation [13]. One EVA replaceable unit had been launched on STS-132 in 2010 and stored on MRM-1 [1]. No held source reports flight values for its accuracy, brake torque or backlash; the tables below give specification and ground qualification values [1], [2].

Chinese Space Station. The core module arm is credited by the operator with 738 kg, 10.2 m, 25 t drag capacity and 0.6 m/s tip speed, without a stated measurement basis [18]. On 6 January 2022 it unlocked the Tianzhou-2 cargo craft, moved it and re-docked it in 47 minutes; the report gives no loads or alignment data [17]. The experiment module arm carries a six-axis force and torque sensor with 32 strain gauges, ranged at plus or minus 500 N and plus or minus 100 N m, whose vibration and thermal vacuum tests are reported as passed without levels [37].

Robot assistants inside the station. Robonaut 2 arrived in February 2011 and ran fixed-base task board experiments for more than two years, with climbing legs and a battery backpack in preparation [20], [21]. The status report is narrative, with no task success rates or cycle counts [21]. Astrobee replaced SPHERES as an operations tool, which had been confined to a two meter cube by its ultrasonic beacons; the pre-launch description contains no measurements [22]. Int-Ball2 made its first on-orbit demonstration on 17 October 2023, and the on-orbit reference marker system itself carries up to 28 mm position error [38]. A joint review compares Astrobee, CIMON and the Int-Ball vehicles side by side [39].

What is being tried now

Joint and brake design for long life. ERA’s joint is the reference design: a four stage 450:1 planetary train with tandem stage pairs on floating ring wheels [1], and a motor unit with look-up-table current shaping that cut torque ripple from about 15 percent to under 0.3 percent peak to peak [2]. It is qualified for a flight arm, not flight-proven. Newer lubricant work compares silahydrocarbons with the incumbent Bray oil on a spiral orbit tribometer and a vacuum four-ball tribometer; that is ground test only [5].

Closed-form and fast inverse kinematics for SSRMS-type arms. A closed-form solution for the offset roll-yaw-pitch geometry of Canadarm2 and the Chinese arms yields sixteen solution sets; it is simulation only, and its two-orders-of-magnitude speed claim is against a general numerical toolbox [32]. A cyclic coordinate descent hybrid reports a 99.32 percent success rate and 4.86 ms solve time on a laboratory arm model, not comparable to the other methods [31]. A decoupling solution for the Chinese Wentian module arm was tested on more than 10,000 simulated poses without hardware [33]. None has flown.

Servicing and debris removal. A tutorial review ties space manipulators to active debris removal, with figures taken from secondary sources [28]. NASA’s ISAM state of play series sorts manipulators by long and short reach around an 8 m discriminator and lists the Chinese core arm at 7 DOF and 10 m [19]. The 2022 edition shares whole paragraphs with its successor and is not an independent source [27]. A review of orbital servicing technology treats vision and force control as the baseline sensing architecture [9], and the Springer handbook chapter places ERA among the orbital manipulator systems that followed SRMS, ETS-VII and JEMRMS [12].

Telerobotics from orbit. KONTUR-2 flew a 2-DOF force feedback joystick on the Russian segment and measured 20 to 30 ms round trip delay to ground robots over the S-band link, against 65 ms and 5 to 15 percent loss on the training link [24]. The joystick delivers 15 N maximum handle force, and a 1.57 Nm/rad maximum stable stiffness is stated without derivation [25]. ETS-VII, flown in 1999, demonstrated manipulator control experiments and on-orbit inertia identification, each from a single trial [26]; the DLR overview of the same period covers ROTEX and the ground-controlled GETEX experiments [10]. Astrobee’s perching arm has a gripper whose handrail detachment force averaged 6.43 N over five bench trials on one prototype [23]. A proposed standard for lunar arm and tool interfaces defines four notional classes by load, power, data rate and accuracy, with no hardware built to them [40].

Ground facilities. Free-floating servicing dynamics are reproduced on hardware-in-the-loop rigs. DLR’s OOS-SIM uses two KUKA KR120 robots at a 4 ms real-time rate with a common workspace of about 23.28 m3 [30]. York University has built a 6 DOF rig with two manipulators and active gravity compensation, documented only by a one-page abstract [29]. A 2024 survey inventories the ground facilities and the arm specifications, all secondary [35].

The numbers

ERA parameter	Value	Source
Total length	11.3 m	[4]
Reach	9.7 m	[4]
Mass	630 kg	[4]
Rated moveable mass	8000 kg	[4]
Maximum payload dimensions	3 by 3 by 8.1 m	[1]
Maximum tip speed	0.10 m/s	[4]
Braking distance at maximum joint speed	0.15 m	[1]
Positioning accuracy, open loop	+/-40 mm	[1]
Positioning accuracy, closed loop	+/-5 mm	[4]
Peak power / standby / hibernation heater	800 W / 420 W / 250 W	[1]
Cameras	4	[3]

Joint motor unit and brake	Value	Source
Motor stall torque	over 1.45 N m	[2]
Motor no-load speed at 30 V	345 rpm	[2]
Motor mass, size	0.750 kg, 120 mm by 29 mm	[2]
Brake torque variation	within +/-22 percent of average	[2]
Emergency stop time	under 3.5 s	[2]
Qualified duty	4380 h, over 10,700 operations	[2]
Running-in before the brake holds	at least 12,000 revolutions in vacuum	[2]
Torque peak with insufficient third-body particles	1.78 N m against a 1.1 N m limit	[2]
Gearbox	four stage planetary, 450:1 (454:1 in the joint hardware paper)	[1], [2]
Brake disc separation	0.25 to 0.30 mm	[1]

Station arm	Figure	Source
Canadarm2 dry mass	1497 kg	[8]
JEMRMS first eigenvalue, 7000 kg payload (design)	0.046 Hz	[16]
Shuttle arm cross coupling, STS-41C	about 22 percent	[14]
Chinese core arm, operator figures	738 kg, 10.2 m, 0.6 m/s	[18]
Tianzhou-2 transposition	47 minutes	[17]
Survey values: ERA 11.3 m, 7 DOF; Canadarm2 17 m	listed specifications	[34]

What is not established

No held source reports ERA flight performance: accuracy, speed and braking are specification values with no qualification test results or error budget [1]. Brake and motor qualification is unit level and predates launch; the recovery time after ambient exposure and the torque peak correction are stated without data, and the coating’s air and vacuum torque cannot be correlated one to one [2]. System-level operating and survival temperatures are not in the held sources. Three-dimensional operation was qualified by simulation, because only planar motion could be tested on the ground [1]. The Chinese arm figures are agency statements [17], [18], and the JEMRMS eigenvalues are pre-build predictions [16]. The inverse kinematics methods above are simulation results with no flight or hardware comparison [31], [32], [33]. Astrobee and Robonaut 2 results are narrative or bench-scale, without task statistics [21], [23].