Compendium

Space Robotics — Orbital Manipulators, Planetary Rovers, On-Orbit Servicing

27 min read

Space Robotics — Orbital Manipulators, Planetary Rovers, On-Orbit Servicing

Robotics in the space environment: orbital manipulators (Canadarm, Astrobee, ETS-VII), planetary surface rovers (Sojourner through Perseverance, Yutu, CE-3/4/5), free-flying inspection robots (Astrobee, Int-Ball), the emerging on-orbit servicing / assembly / manufacturing (OSAM) sector (MEV-1, OSAM-1, ClearSpace-1). The hostile environment — vacuum, plasma, radiation, ±150 K thermal swings, no GPS off-Earth — drives every design choice away from terrestrial best practice.

See also

1. At a glance

Space robotics is what robotics has to become when you remove the terrestrial assumptions: 1 g gravity, a benign 250–300 K thermal environment, atmospheric convection, GPS, predictable lighting, 100 Mbit Ethernet, and the option to walk over and reboot. Every mission since Lunokhod-1 (1970) has fought four overlapping problems:

  • Radiation — single-event upsets in unhardened CMOS within hours of LEO transit through the South Atlantic Anomaly; total-ionizing dose accumulates over years; heavy ions cause latchup and burnout.
  • Thermal — no convective cooling; ±150 K diurnal swing in LEO; deep-cold shadows; hot Sun side; radiative-only heat rejection.
  • Vacuum — lubricant out-gassing; cold-welding of bare metals; no aerodynamic drag for stabilization; plume impingement during proximity ops.
  • Latency — 1.3 s one-way to the Moon; 4–24 minutes to Mars; no two-way teleoperation across that gap; ~30–80 minutes to the outer planets.

The robot is then one of three operational classes:

  • Crewed-vehicle manipulators — Canadarm / Canadarm2 / Dextre on the ISS, JEMRMS on Kibo, the European Robotic Arm (ERA) on the Russian segment. Payload movers and EVA assist, teleoperated by astronauts or ground.
  • Planetary surface rovers — Mars: Sojourner (1997), Spirit + Opportunity (2004), Curiosity (2012), Perseverance (2021), plus ESA’s stalled ExoMars Rosalind Franklin. Moon: Lunokhod-1/2, Yutu-1/2, Chang’e 5 sampler, ISRO Pragyan (2023), JAXA SLIM (2024). Sojourner-class rovers move meters/sol; Perseverance averages ~100 m/sol with AutoNav.
  • On-orbit servicing (OOS / OSAM) — Northrop Grumman’s MEV-1 (2020) and MEV-2 (2021) docked with Intelsat satellites for life-extension; NASA’s OSAM-1 (cancelled 2024 but concept lives at SPACE-X, Orbit Fab, Astroscale); DARPA’s RSGS for GEO refuel/repair; ESA’s ClearSpace-1 for active debris removal of the VESPA upper stage.
  • Free-flyers and proximity-ops drones — NASA’s Astrobee fleet on the ISS (Bumble, Honey, Queen, since 2019); JAXA’s Int-Ball / Int-Ball 2 (2017 / 2024) for video documentation; Ingenuity Mars Helicopter (2021–2024, 72 flights).

Where this sits.

  • ADCS supplies host pointing.
  • Orbital mechanics supplies relative-motion equations (Clohessy-Wiltshire for proximity ops).
  • Serial-arm kinematics applies, but the base is free-floating — the Jacobian includes base reaction.
  • Teleoperation handles the human-in-the-loop with predictive display compensating for latency.
  • Safety standards are NASA-STD-8719.13C (software) and NASA-HDBK-8719.14 (mission risk classification) rather than ISO 10218.
  • Power is solar + Li-ion or RTG, not wall-plug.
  • SLAM runs on rad-hard processors at single-Hz with strict CPU/RAM caps.

First ask before applying:

  • What is the radiation environment? LEO behind 100 mil Al: 1–10 krad/yr. GEO: 10–50 krad/yr. Interplanetary: GCR-driven SEU rates 10⁻⁷ to 10⁻⁵ upsets/bit-day. Rad-hardened-by-design (RHBD) parts cost 10–100× commercial.
  • Is the manipulator base fixed or floating? Fixed (ADCS holds inertial) → fixed-base with disturbance rejection. Floating → generalized Jacobian (Vafa-Dubowsky 1987) accounting for base reaction.
  • Round-trip latency? < 0.5 s → direct teleop. 1–10 s → predictive display + supervisory. > 60 s → autonomous with intent specification.
  • Contact involved? Berthing/grapple = high-force; respect structural limits; use soft-capture mechanisms (NASA Docking System NDS, IDSS standard).
  • What’s the qualification regime? Crewed = NASA-STD-3001 + STD-8719.13 two-fault tolerance. Uncrewed = mission-class A (highest, e.g., Mars 2020) → mission-class D (lowest, e.g., CubeSats).

2. First principles

2.1 Free-floating manipulator dynamics

A manipulator on a free-floating base (e.g., a satellite-servicer with thrusters off to save fuel) is a floating-base multi-body system. The equations of motion in joint-space form follow rigid-body dynamics but the 6-DoF base has no actuation:

H(q) q̈ + C(q, q̇) q̇ = [τ_base; τ_joints]

with τ_base = 0 when thrusters are off. Conservation of total linear and angular momentum then constrains q̈:

Σ m_i ṙ_i = p₀ (conserved) Σ (m_i r_i × ṙ_i + I_i ω_i) = L₀ (conserved)

The generalized Jacobian matrix J_g(q) (Umetani & Yoshida 1989, Vafa & Dubowsky 1987) maps joint rates to end-effector twist while respecting momentum conservation:

v_ee = J_g(q) q̇_joints

with the implicit base motion absorbed into J_g. J_g depends on the instantaneous configuration, not just joint angles — moving the arm out at q₁ = 90° gives different base reaction than at q₁ = 0°. Singularities in J_g are called dynamic singularities and have no analogue in fixed-base robotics; they trap the end-effector at configurations where no joint motion can produce certain Cartesian motions.

2.2 Reaction-null space (RNS)

The set of joint motions q̇ for which the base does not react. Yoshida 1995 showed that for a non-redundant arm RNS is empty; for an n > 6 DoF arm (the Canadarm has 7 DoF for exactly this reason) RNS is (n-6)-dimensional. Operating the arm inside RNS preserves carrier attitude — important when the carrier is also pointing antennas, solar panels, or science instruments.

2.3 Relative motion in orbit (CW equations)

For a chaser in close proximity (< 1 km) to a target in a near-circular orbit of mean motion n, the Clohessy-Wiltshire (or Hill-CW) equations describe the chaser’s position (x, y, z) in the target’s local-vertical-local-horizontal (LVLH) frame, with x along radial, y along velocity, z out-of-plane:

ẍ − 2 n ẏ − 3 n² x = a_x ÿ + 2 n ẋ = a_y z̈ + n² z = a_z

Closed-form solutions exist; the V-bar approach (along ±y from below) and R-bar approach (along ±x from below) are the two standard approach corridors. SpaceX Dragon, Cygnus, and HTV all use V-bar; Soyuz uses an inertial KURS rendezvous. R-bar is preferred for tumbling targets because it requires less plume impingement on the target.

2.4 Radiation effects

Three classes:

  • Total Ionizing Dose (TID): cumulative damage from electrons + protons trapped in Van Allen belts. Threshold for COTS CMOS: 5–20 krad(Si). Rad-hard parts: 100–1000 krad(Si). Measured in rad(Si) or Gy.
  • Single-Event Effects (SEE): instantaneous upsets from heavy ions. Single-Event Upset (SEU) flips a bit in SRAM/registers; Single-Event Latch-up (SEL) shorts power; Single-Event Burnout (SEB) destroys MOSFETs. Triple Modular Redundancy (TMR) + scrubbing handles SEUs in FPGAs; current-limiting + auto-power-cycle handles SELs.
  • Displacement Damage: cumulative lattice damage in bipolar transistors and CCDs from non-ionizing energy loss (NIEL). Drives CCD dark-current up over time; mitigated by cooling and by switching to CMOS image sensors which are inherently more tolerant.

Rad-hard processors used in flight: RAD750 (PowerPC-derived, Curiosity / Perseverance / Webb / Mars-2020), RAD5545, LEON3FT/4 (SPARC-based, used in ExoMars + ESA missions), and the emerging high-performance class — BAE RAD5500, Microchip PIC32MZ-DA, Cobham Gaisler GR740.

2.5 Thermal in vacuum

No convection; only conduction (along structure) and radiation (to deep space). Equilibrium for a flat absorbing surface in Sun-illumination at 1 AU:

α_s · S = ε · σ · T⁴

with S = 1361 W/m² (solar constant at 1 AU), σ = 5.67×10⁻⁸ W/m²·K⁴. For α_s/ε = 1 (bare aluminum oxidized) and S = 1361, T_eq ≈ 393 K (120 °C). For α_s/ε = 0.2 (white paint), T_eq ≈ 271 K. In eclipse the same surface radiates to ~3 K background → equilibrium ~ 0 K, except for internal heat dissipation and heater inputs. Robots use heaters, multi-layer insulation (MLI, 10–30 layers of aluminized Mylar/Kapton), and survival heaters to keep electronics > -40 °C and lubricants > -55 °C during night.

2.6 Tribology in vacuum

Cold-welding: two clean metals in contact in vacuum bond at the contact patch within minutes. Mitigated by dissimilar metals, hard coatings (TiN, DLC), and dry-film lubricants. Conventional liquid greases out-gas, contaminating optics. Standard space lubricants: Krytox 143AC / Braycote 601EF (PFPE oils with low vapor pressure, < 10⁻¹³ Torr at 25 °C), MoS₂ (dry-film, 0.04 friction coefficient in vacuum, but degrades in humid air during ground test), PTFE-based composites for plain bearings. Bearings use solid-film MoS₂ on rolling elements + sputtered Pb on cages.

2.7 Wheel/regolith interaction (terramechanics)

For planetary rovers driving on lunar or Martian regolith — fine, low-bulk-density granular media. Wheel sinkage z is governed by Bekker’s pressure-sinkage equation (Bekker 1969):

p = (k_c / b + k_φ) · z

with b = wheel width, n ≈ 1.0–1.2 for fine regolith, k_c and k_φ empirical constants. Drawbar pull DP = T_wheel − R_resistance. Wong 1989 derived the full slip-traction curves. Mars rover wheels (rocker-bogie) use grouser bars (transverse ribs) for shear traction; Curiosity’s wheels famously wore through punctures from sharp rocks because the 0.75 mm Al skin wasn’t designed for the rock distribution at Gale crater.

2.8 Rocker-bogie kinematics (passive equalization)

Six-wheel passive suspension used by every NASA Mars rover (Sojourner through Perseverance):

  • Two side-mounted rockers (front + middle wheel) pivot at chassis.
  • Two side-mounted bogies (middle + rear wheel) pivot at the rocker.
  • Differential bar (or cable, in some designs) couples the two rockers so the body tilts at half the average rocker angle.
  • No springs — pure rigid-body geometric averaging.
  • Result: chassis tilt halved relative to terrain irregularities; all six wheels maintain ground contact over obstacles up to ~ wheel-diameter without losing traction.

Sinkage equalization implies static-stability at climbing angles up to ~45° (Mars rovers driven at < 30° per safety policy). Wheel-walking modifications (Lunokhod, ExoMars) enable additional thrust by sequentially actuating wheels.

2.9 Stereo visual odometry for rovers

Standard pipeline:

  1. Acquire stereo pair from Navcams (Mars: 2 cameras 30 cm apart on mast).
  2. Detect features (FAST corners on Curiosity; ORB on Perseverance + Yutu).
  3. Match left-right within frame for stereo depth.
  4. Match frame-to-frame for motion.
  5. RANSAC + minimization of reprojection error → 6-DoF rover pose delta.
  6. Concatenate to wheel-odometry to get robust pose estimate.

Wheel odometry alone has slip error 10–40%. Visual odometry: 1–2% over 10 m. After 100 m the drift accumulates to several meters; loop-closure via orbital images (Mars Reconnaissance Orbiter HiRISE) closes the global drift.

2.10 Mass moment cancellation in OOS

When a chaser docks with a target, the post-docking system has combined inertia I_total ≈ I_chaser + I_target. ADCS sized for I_chaser alone is now under-sized. Two responses:

  • Soft-docking — passive damping mechanism absorbs the kinetic energy without significantly perturbing the combined system. NDS standard provides ~1000 N·s damping.
  • Reaction-wheel saturation early — chaser wheels saturate; thrusters must take over. Pre-plan dump trajectory in advance.

OOS architecture must include propellant budget for the post-docking attitude-stabilization burn(s), not just for the rendezvous.

3. Practical math — orbital and surface ops

3.1 Sized example: ISS robotic arm reach

Canadarm2 (SSRMS): 17.6 m tip-to-tip, 7 DoF (3-1-3 wrist), 1626 kg, can move 116 t payloads in microgravity. Max tip speed for unloaded operation: 0.37 m/s (single joint); for loaded: 0.02 m/s. Joint torque budget at the largest joint (shoulder pitch): ~3000 N·m. The Latching End Effector (LEE) at each end is interchangeable (you can “inchworm” along the truss by alternating which end is grappled).

3.2 Sized example: Mars rover sol-budget

Perseverance averages: drive ~100–200 m, science ops 2–4 hr, communication windows 2× per sol via MRO/MAVEN/Trace Gas Orbiter (UHF, 2 Mbps for ~6 minutes per pass). Power budget ~110 W continuous from MMRTG (multi-mission RTG, 4.8 kg Pu-238, ~2 kW thermal, ~110 W electrical, degrading ~2%/yr). Battery (43 A·h Li-ion × 2) holds the night load. Drive energy: ~25 Wh per meter (compared to ~3 Wh/m for Curiosity’s lighter chassis).

3.3 Plume impingement

Thruster plumes during proximity ops impinge on the target → exert force/torque + contaminate surfaces. Plume model: free molecular flow, Knudsen number > 10, density falls as 1/r² with cosine angular distribution. Standard mitigation: stop thrusting beyond a “keep-out zone” inside ~30 m and coast inertially, or use cold-gas / electric propulsion with lower mass flow.

3.4 AutoNav driving cycle (Perseverance)

The driving loop on Mars when in AutoNav:

  1. Acquire stereo pair from Navcams (1 Hz cycle ground processed; 0.2 Hz onboard).
  2. Compute disparity → point cloud (rad-hard FPGA: Xilinx Virtex-5QV).
  3. GESTALT terrain assessment (NASA JPL) classifies each 20×20 cm cell as drive/sense-hazard/keep-out.
  4. Field D-star plans a path through the cost map (Stentz 1994 / Ferguson & Stentz 2006).
  5. Execute 1–3 m segment.
  6. Visual-odometry (FAST features tracked across frames) checks slip; if slip > 40% → halt.
  7. Repeat.

Closed-loop speed: 0.04–0.12 m/s including stops. Open-loop “blind drives” (when terrain ahead is human-vetted) reach 0.15 m/s.

3.5 EVA-style berthing torque budget

Canadarm2 (SSRMS) limits during ISS payload moves:

  • Tip speed unloaded: 0.37 m/s.
  • Tip speed loaded (10 t+): 0.02–0.05 m/s.
  • Maximum tip force in contact: ~445 N (~100 lbf), limited by braking system.
  • Maximum tip torque: ~70 N·m.
  • Soft-stop on contact-detect: 0.3 s ramp to zero.

For Dragon berthing: approach speed at capture < 0.04 m/s; capture latency < 0.5 s; SPDM “Dextre” then takes over for fine manipulation (capable of swapping ORUs to ~mm precision).

3.6 Battery sizing for Mars night

Perseverance battery (2× 43 A·h Li-ion @ 28 V nominal) sized for:

  • MMRTG output ~110 W continuous (degrading 2%/yr Pu-238 decay).
  • Night load: heaters + comms = ~100 W avg; peak transient 200 W.
  • Sol cycle 24h 39min; night ~12 h.
  • Required night Wh = 100 W × 12 h = 1200 Wh; battery capacity = 43 × 28 = 1200 Wh × 2 = 2400 Wh.
  • Margin: 50% (one battery only as backup).
  • DoD: < 60% to preserve cycle life over multi-year mission.

3.7 Hot-cold thermal cycling

For a lunar surface robot at the equator:

  • Day equilibrium (full sun, α/ε = 0.3 painted): T ≈ 360 K.
  • Night equilibrium (no sun, radiating to 3 K, internal Q = 20 W): T ≈ 130 K.
  • Cycle duration: 14 Earth days / 14 Earth days.
  • Required heater power at night to keep electronics > 240 K: ~50–80 W.
  • ESA’s Lunar Rover Concept (Pangea-X / Heracles): RHU (Radioisotope Heater Unit, 1 W each, multiple) + electrical heaters powered from solar + battery.

4. Design heuristics

  • Always assume the part will see SEU. Even rad-hard parts upset. Design with watchdogs at every level. Scrub FPGA configuration memory every 100 ms minimum.
  • Two-fault tolerance for crewed missions. NASA-STD-3001 + STD-8719.13: any two independent failures must not cause loss of crew. Adds redundancy weight (3–5×) compared to single-fault tolerance.
  • Heritage trumps performance. Components flown before (PPL Class-1 / S-class EEE parts) cost 10× more but reduce qualification risk to near-zero. New-flight-experience parts often slip launches.
  • Mass budget is law. 80k per kg to LEO, $1M+/kg to Mars surface. Every gram of structure justifies engineering hours.
  • No-touch reachability matters. A rover stuck in soft sand (Spirit, 2009, sol 1899) cannot be physically rescued. Path planners must reason about slope, slip risk, and 1-way return clearance.
  • Don’t trust a single sensor. Spirit’s right-front wheel failed in 2006; her remaining 5 drove backwards for the rest of the mission. Build mission concepts that degrade gracefully.
  • Latency drives architecture. With one-way Mars latency of 5–22 minutes you cannot teleoperate. With one-way Moon latency of 1.3 s you can — Lunokhod-1 was driven that way for 322 days.
  • Cold-soak everything. Lubricants, batteries, and electronics get qualified to the survival temperature range (typically −55 to +85 °C operational, −80 to +125 °C survival).
  • Test as you fly. Hardware-in-the-loop with thermal-vacuum chamber + radiation source for piecewise verification before integrated environmental test.
  • Margin is the answer to “what if?“. Mass margin 20% at PDR, 10% at CDR. Power margin 30% at PDR. Schedule margin 30% (NASA standard); always burnt by the launch slip.
  • Plan for “off-nominal” first. Most flight software is for safe-mode + fault recovery, not nominal operations.
  • Single-string only for risk class C/D. Class A/B requires redundant strings (computer, battery, comm) with cross-strap.
  • Use space-qualified parts lists. ESA QPL (Qualified Parts List), NASA EEE Parts List, JAXA HK list. Substitutions require requalification.

5. Components & sourcing

SubsystemStandard parts
Rad-hard CPUBAE RAD750 (Curiosity, Perseverance, Webb), RAD5545 quad-core, Cobham Gaisler GR740 (LEON4FT quad-core, ESA flagship)
Rad-tolerant FPGAMicrochip RT PolarFire, Xilinx XQRKU060 (Kintex UltraScale rad-tolerant), Xilinx Virtex-5QV (Perseverance vision)
EEE partsMil-PRF-38535 (microcircuits), S-class screening, JANS for transistors, ESCC-QPL for ESA
Solar panelsSpectrolab XTJ-Prime (multi-junction GaAs, ~30% efficient, 10k per cell), AzurSpace 4G32C
BatteriesSaft VES16 / VES180 Li-ion (300+ Wh/kg, qualified to 30k+ LEO cycles), Eagle-Picher LSB-series
Reaction wheelsHoneywell HR-12 / HR-16, Rockwell Collins (now Collins Aerospace) M-series, Bradford Engineering W18
Star trackersSodern Auriga (10 arcsec, 5 W), Jena-Optronik ASTRO APS, Ball Aerospace HD-1003
IMUHoneywell HG9900, Northrop Grumman LN-200S, Safran Scorpio-CM4
LiDARVelodyne (custom space variants for ATV/HTV), Optech ILRIS (used on Demonstrator for Autonomous Rendezvous Technology DART, 2005, which famously crashed into its target)
ThrustersAerojet MR-103 (1N cold gas), Moog Bradford cold-gas micro-Newton, Busek BHT-200 Hall-effect
Robotic jointsCustom (no COTS) — MDA (Canadarm), Honeybee Robotics (Sample Caching System), MMA Design
Solar absorbersAZ Technology Z-93 white paint, OSR (optical solar reflector) tiles

5.1 Common space-robotics software frameworks

- F' (F-Prime) — NASA JPL flight software framework (C++); Mars 2020 / Ingenuity
- cFE / cFS (Core Flight Software) — NASA Goddard; reusable middleware
- NASA OS Abstraction Layer (OSAL) — portable across VxWorks / RTEMS / Linux
- ROS 2 + space2 / Astrobee variants — for ISS-grade robots (Astrobee uses ROS Kinetic legacy)
- ESA Tasking Framework (TasC) — used on ExoMars, BepiColombo
- TASTE (ESA) — model-driven dev for embedded space systems
- Robotic Operating System for Space (ROS-Space, ESA experimental)
- Custom: most flight programs roll their own task scheduler atop VxWorks or RTEMS RTOS

5.2 RTOS landscape

- VxWorks (Wind River) — Curiosity, Perseverance, Webb, most NASA missions
- RTEMS (open source) — ESA preferred; many CubeSats; Ingenuity uses Linux + Snapdragon
- FreeRTOS — CubeSat workhorse
- INTEGRITY-178 (Green Hills) — DO-178C certified, military space
- LynxOS-178 — military space
- Linux (Yocto/Buildroot custom) — Ingenuity, modern computer-vision payloads

6. Reference data

6.1 Notable space-robotics missions

MissionYearClassNotes
Lunokhod-1/21970/1973Lunar roverFirst teleoperated planetary rover; 10.5 km / 39 km traverse
Sojourner (Pathfinder)1997Mars rover11.5 kg, 100 m total, demonstrated rocker-bogie
ETS-VII (Engineering Test Satellite 7)1997OOS demoFirst on-orbit servicing demo; Japanese 2 m manipulator + autonomous rendezvous
Spirit + Opportunity2004Mars rovers (MER)174 kg each; designed for 90 sols, lasted ~7 years (Spirit) and ~15 years (Opportunity)
Canadarm2 / Dextre2001 / 2008ISS manipulator17.6 m, 7 DoF + 15-DoF SPDM “Dextre” for fine ops
Curiosity2012Mars rover (MSL)899 kg, MMRTG-powered, ~32 km traversed by 2026, Sample Analysis at Mars instrument
Chang’e 3 / Yutu-12013Lunar rover (China)First lunar lander since 1976; Yutu-2 active until 2024
Perseverance + Ingenuity2021Mars rover + helicopter1025 kg rover, 1.8 kg helicopter (72 flights, 17 km total before motor failure Jan 2024)
MEV-1 / MEV-2 (Northrop Grumman)2020 / 2021OOS commercialDocked with Intelsat 901 / 10-02 for 5-yr life extension
Chandrayaan-3 / Pragyan2023Lunar rover (India)South pole region; first lunar south-pole soft landing
SLIM (JAXA)2024Lunar lander”Pinpoint” landing accuracy 100 m, two micro-rovers
Astrobee (NASA)2019–Free-flyerThree ISS units; perch arm; vision-based localization via NavCam + Dock recharge

6.2 Thermal extremes

BodyDay tempNight tempNotes
LEO-150 to +120 °C(orbital eclipse cycle)16 cycles/day at 90-min orbit
Lunar surface+127 °C-173 °C (polar shadow: -243 °C)Permanently-shadowed regions: -250 °C
Mars equator+20 °C-73 °CDiurnal swing on rover deck
Mars pole-90 °C-125 °CPhoenix lander could not survive winter
Europa-160 °C-220 °CRadiation belt 100× LEO

6.3 Surface gravity reference

Bodyg (m/s²)g/g_earth
Earth9.811.000
Moon1.620.165
Mars3.710.379
Mercury3.700.378
Venus8.870.905
Jupiter (cloud)24.792.529
Europa1.310.134
Titan1.350.138
Ceres0.270.028
Phobos0.00570.00058

Lower gravity changes everything: hopping easier, jumping easier, but anchoring harder (Philae bounce 2014, OSIRIS-REx sample anchoring).

6.4 Communication delays

LinkOne-way latencyBandwidth
ISS ↔ Earth (TDRSS)< 0.3 sup to 600 Mbps
Moon ↔ Earth1.28 s100 Mbps (LRO LRO-comm)
Mars ↔ Earth4–24 minutes2 Mbps (UHF via orbiter relay), 0.5–2 Mbps direct X-band
Outer planets30–80 minutes< 100 kbps (Voyager)

7. Failure modes & debugging

  • SEU-induced fault — Cassini’s solid-state recorder dropped bits during Saturn flyby. Fix: scrub + ECC; reset processor on watchdog timeout.
  • Cold-welded mechanism — Galileo high-gain antenna failed to deploy (1991), attributed to bonded sleeves + ribs after long thermal soak. Fix: heaters + repeated motion + plasma-clean before launch.
  • Lubricant migration — Hubble fine-guidance sensor degradation traced to PFPE breakdown in vacuum. Fix: switch to MoS₂ for high-cycle bearings.
  • Wheel damage — Curiosity wheels punctured by Gale crater rocks. Fix: rover-driving team adopted “Traction Control” with 6-wheel slip monitoring + path planner that prefers smooth terrain.
  • Bit-flip in star-tracker firmware — Mars Global Surveyor lost in 2006 after sequence-of-events caused battery overheating; root cause traced to a year-old uplinked parameter table not properly version-checked. Fix: bidirectional checksums on every uplink, archive of “as-flying” tables, simulation in twin facility before commanding.
  • Anomaly in autonomous rendezvous — DART (2005) collided with target Mublcom after relative-nav diverged. Fix: independent absolute-nav reference (GPS works in LEO; ground tracking for higher orbits).
  • Cosmic-ray-induced power transient — Voyager 2 attitude excursion 2010 attributed to an upset in the flight data subsystem. Fix: rebooted with safe-mode config; data routing reconfigured.
  • Sun-keep-out violation — Hubble fine-guidance sensors damaged by Sun acquisition glitch. Fix: software-enforced Sun-exclusion cones; multi-stage filter on attitude commands.
  • Tribological seizure — Mars Climate Orbiter joints (had it survived EDL) and Mars Polar Lander (likely cause of failure) — frozen actuators after long cruise. Fix: pre-cruise lubricant cycling regimen; mid-cruise heater check-outs.
  • Stale onboard ephemeris — DART again; the relative ephemeris of Mublcom was off by km because the prediction window was too coarse. Fix: bidirectional updates; conservative ground-update before any close-approach.
  • GNC software requirements mismatch — Mars Climate Orbiter (1999) — Lockheed Martin used pound-force, JPL used Newtons; the trajectory burn was off by a factor 4.45. Fix: explicit unit standards in interface control documents; automated unit-checking in build chain.
  • Power-bus contactor failure — Hayabusa 1 had multiple thruster losses, fuel leak, communication outage; recovered after 4 years. Fix: redundant strings + ground-based extraordinary recovery operations.
  • Reaction-wheel friction increase — Kepler space telescope lost wheels 2012 + 2013 after dust contamination of bearings. Fix: K2 mission re-aimed the telescope to balance pressure of sunlight against remaining 2 wheels.

8. Case studies

8.1 Canadarm2 berthing of SpaceX Dragon

Dragon approaches ISS along the R-bar (below) using GPS-relative nav + LIDAR (CDGPS internal, RGPS to ISS). At ~10 m it captures station-keeping. Astronaut at the Cupola Robotic Workstation uses Canadarm2 (7-DoF) with two hand-controllers (rotational + translational) to drive the SSRMS’s LEE to Dragon’s grapple fixture, captures (~30 s drift toward Dragon), and inchworms back to a berthing port on Harmony or Unity. Snares engage at ~0.04 m/s. Mass moved: ~12 t. The whole sequence takes ~45 min from “approach complete” to “Dragon hard-mated.”

8.2 Perseverance landing via Sky Crane + first 100 sols

EDL (Entry, Descent, Landing) — atmospheric entry at 5.4 km/s, supersonic parachute at Mach 2, then a powered descent stage (“Sky Crane”) lowering the rover on three nylon bridles while hovering at 20 m, finally cutting the bridles after touchdown. Terrain Relative Navigation (TRN) compared real-time descent imagery against orbital maps to retarget the landing ellipse from 6 km × 7 km to ~1 km. After landing, surface ops began: deploy mast, image surroundings (Mastcam-Z), check radio links, deploy helicopter Ingenuity from underside (April 19, 2021 first flight, 39 s, 3 m altitude), then drive ~5 m to clear the landing site and begin sampling at Jezero crater.

8.3 MEV-2 docking with Intelsat 10-02

Northrop’s Mission Extension Vehicle 2 launched August 2020, rendezvoused with Intelsat 10-02 at GEO in April 2021 using a forward-mounted docking probe + visual inspection cameras. It engaged the customer satellite’s apogee-engine throat (a docking interface not designed for that purpose, exploited by MEV’s probe + soft-capture grippers). MEV-2 now provides station-keeping + attitude control to extend Intelsat 10-02’s mission by 5 years. MEV-1 docked with Intelsat 901 in February 2020 — first commercial GEO life-extension docking. The OSAM-1 follow-on (NASA Glenn) was canceled in 2024 due to budget overruns; the commercial mission concept survived in newer SPACE-X / Orbit Fab / Astroscale programs.

8.4 Ingenuity Mars Helicopter

Stanford / NASA JPL collaboration. 1.8 kg, two counter-rotating 1.2 m carbon-fiber rotors at 2400 rpm (Mars atmosphere has 1% Earth density → must spin 10× faster than a terrestrial heli of same size). Powered by a 35.75 Wh Li-ion pack + solar recharge over Mars sols. 6× MMU MEMS IMUs, 5 MP downward camera + 13 MP nav camera, Snapdragon 801 (yes, the cell-phone processor) for vision processing, FPGA for flight control. Algorithms: visual-inertial-odometry estimating velocity at 500 Hz. Flew 72 times (April 2021 – January 2024), covered 17 km, reached 24 m altitude, before a rotor-blade tip strike on landing damaged the rotor. Mission ended Jan 25, 2024.

8.5 Astrobee on ISS

3 free-flyers (Bumble, Honey, Queen) inside the ISS USOS modules. ~30 cm cube, electric ducted-fan propulsion (4 nozzles → 6-DoF), perch arm (one-DoF jaw to grab handrails for stand-by power), three Snapdragon CPUs (LLP, MLP, HLP), NavCam (vision-based localization against an a priori 3D map of the module interiors built from Astrobee SLAM), and a dock for autonomous recharge. Used for: assisting astronauts (carry tools/cameras), recording video, hosting guest-science payloads. Open-source software stack (ROS-based), released as Astrobee Robot Software on GitHub.

8.6 ETS-VII (Japan, 1997) — first OOS demo

JAXA’s Engineering Test Satellite-7: first satellite to autonomously dock + perform robotic-arm tasks in space. Two-vehicle system — “Hikoboshi” target + “Orihime” chaser — with the chaser hosting a 2 m, 6-DOF manipulator. Demonstrated: autonomous rendezvous + docking from 9 km, ORU exchange, drilling, electrical connector mate, truss assembly. Operated 1997–1999, returning the test data that informed JEMRMS design + the entire modern OOS literature.

8.7 ExoMars Rosalind Franklin — design + status

ESA + Roscosmos collaboration; Roscosmos withdrew 2022 due to Russian invasion of Ukraine. Rover (~310 kg) designed to drill 2 m into Martian subsurface searching for biosignatures. Six-wheel rocker-bogie with “creeping” gait (wheel-walking) for soft terrain. NavCam + PanCam + WISDOM ground-penetrating radar. Currently re-planned for launch 2028+ with NASA EDL support (re-aim from Lavochkin Russian lander to NASA-built skycrane-style EDL).

8.8 Hayabusa 2 sample return (JAXA, 2014–2020)

Asteroid Ryugu sample return. Robotics aspects:

  • Sampler horn: cylindrical tube touched asteroid surface at < 0.1 m/s; fired projectile (5 g, 300 m/s) into surface; captured ejecta in catcher.
  • Small Carry-on Impactor (SCI): 2 kg shaped charge fired 14 m/s impactor; created artificial crater for second sample of subsurface material.
  • MINERVA-II hoppers: three small ~1 kg rovers (Rover 1A, 1B, MASCOT) hopped across surface in microgravity using internal torquer reaction.
  • Touch-and-go autonomous control: target marker placed by chaser; chaser navigated via LIDAR + ONC-T camera vision feedback to within 50 cm of marker.

Sample returned to Earth December 2020 (5 g of Ryugu material). Set the architectural template for similar OSIRIS-REx (NASA, Bennu, 2023 return) and the planned MMX (JAXA, Phobos, 2026 launch).

8.9 Lunokhod-1 — the first teleoperated planetary rover

USSR, 1970. ~756 kg, 8 wheels, nuclear-isotope heat source (Po-210 for night warmth), solar-powered driving. Operated by a 5-person ground team in Crimea: commander, driver, navigator, engineer, radio-operator. 1.3 s round-trip latency just barely tolerable for a “watch + drive” loop — the team developed a forecasting / waypoint protocol where the driver would issue a heading + speed + duration command, then wait to see what happened. 322 days of operation, 10.5 km traverse, 25,000 images. Set the precedent for every subsequent rover including Mars architectures.

8.10 OSIRIS-REx sample collection at Bennu (NASA, 2020)

Goddard / Lockheed Martin collaboration. Asteroid Bennu sample-return mission. Robotic aspects:

  • TAGSAM (Touch-And-Go Sample Acquisition Mechanism) — articulated arm extending 3.4 m with a sampler head; nitrogen-blow technique to lift surface material into catcher.
  • Touch-and-go (TAG) maneuver: spacecraft descended to surface at 0.10 m/s, contacted Bennu, fired N₂ canister, lifted off all within < 5 s of contact.
  • OLA (OSIRIS-REx Laser Altimeter) + PolyCam / NavCam / SamCam for terrain-relative navigation.
  • Bennu’s surface turned out to be unexpectedly loose (rubble pile, < 1% gravity) — the sampler sank > 0.5 m before lift-off; sample mass over-flowed the container and a “Mylar flap” failed to close fully. Recovered ~250 g of sample (vs. planned 60 g).
  • Sample returned to Utah desert September 2023.

Influenced subsequent mission designs (Hera, M-MX, Apophis Explorer) toward more conservative anchoring + sampling architectures.

8.11 ClearSpace-1 — first active debris removal (ESA, 2026+)

ClearSpace SA (Swiss startup) + ESA. Mission to deorbit the VESPA upper-stage section left from Vega’s 2013 flight. Approach:

  • 4-arm robotic capture system folds around the target object.
  • Target is ~100 kg, tumbling, no docking interface, never designed to be captured.
  • Mission profile: rendezvous, characterize tumble, match rotation, grapple, perform combined-deorbit burn.
  • Originally planned 2025 launch; now slipped past 2026 due to funding + technical risk.
  • Significance: first non-cooperative target capture in space; precedent for the broader space-debris-removal industry (Astroscale, Northrop Grumman).

8.12 Astrobee NavCam-Localization (Bualat 2018)

Astrobee’s localization architecture:

  • A priori map of ISS USOS modules built from prior Astrobee surveys.
  • 1280×960 NavCam at 1–5 Hz.
  • ORB feature extraction onboard.
  • Pose-graph optimization with prior map as fixed-anchor constraint.
  • Result: ~5 cm position accuracy, ~5° orientation, in real-time on Snapdragon 805.

The Astrobee software stack is open-source on GitHub (Apache 2 license), making it the most accessible space-robot codebase for research labs.

Adjacent

Standards & qualification environments

Radiation testing:
  - Co-60 gamma source (Total Ionizing Dose)
  - Heavy-ion cyclotron (LBNL 88-inch, BNL 6 MV Tandem, Brookhaven NSRL) for SEE
  - Proton beam at IUCF / Tri Lab / Paul Scherrer for displacement damage + SEE
  - Neutron beam at LANSCE for displacement damage

Thermal qualification:
  - Thermal-vacuum chamber (TVAC) at JPL, GSFC, ESTEC, JAXA Tsukuba
  - Solar simulators with Xenon-arc lamps for diurnal cycling
  - Cryogenic test (LN2 shroud) for cold survival

Vibration / mechanical:
  - Sine + random vibration tables (e.g., Sandia STL)
  - Acoustic chamber (HEAT facility, NASA GSFC)
  - Shock test (pyrotechnic shock)
  - Sine-burst (rapid sweep) for resonance survey

Vacuum + outgassing:
  - High-vacuum bake-out (10⁻⁶ Torr at 50–70 °C, 24 hr+)
  - TML (Total Mass Loss) + CVCM (Collected Volatile Condensable Materials)
    per ASTM E595 / ECSS-Q-ST-70-02C — required < 1% TML, < 0.1% CVCM

EMC / EMI:
  - MIL-STD-461F / -462 conducted + radiated emissions + susceptibility
  - ECSS-E-ST-20-07C electromagnetic compatibility

Open-source space-robotics resources

  • F’ (NASA JPL, github.com/nasa/fprime) — flight software framework used on Mars 2020.
  • cFE/cFS (NASA GSFC, github.com/nasa/cFS) — core flight system.
  • OpenMCT (NASA Ames) — mission-ops dashboards.
  • Astrobee Robot Software (NASA Ames, github.com/nasa/astrobee) — full ISS free-flyer stack.
  • NASA-LIS (Lunar Information System) reference data.
  • NAIF SPICE Toolkit (JPL) — ephemeris + frame transforms; Python + Java + C + Fortran wrappers.
  • OreKit (ESA-tied open project) — Java orbit propagation library.
  • GMAT (General Mission Analysis Tool) (NASA Goddard) — open-source mission design.

Mission-class summary (NASA risk classes)

Class A:  Highest priority (e.g., crewed missions; Mars 2020, JWST).
          Two-fault tolerance. Full ground-test program. Cost: $1B+.

Class B:  High priority (e.g., New Horizons, Cassini).
          Single-fault tolerance with critical-item redundancy. Cost: $500M-$1B.

Class C:  Medium priority (e.g., Discovery-class missions).
          Selective redundancy. Cost: $100M-$500M.

Class D:  Lower priority (e.g., CubeSats, Hayabusa-2 sub-missions).
          Often single-string. Higher acceptance of risk. Cost: <$100M.

Citations

  • Vafa Z., Dubowsky S. (1987) “The Kinematics and Dynamics of Space Manipulators: The Virtual Manipulator Approach.” Int J Robotics Research 9(4).
  • Umetani Y., Yoshida K. (1989) “Resolved Motion Rate Control of Space Manipulators with Generalized Jacobian Matrix.” IEEE Trans Robotics & Automation 5(3).
  • Yoshida K. (1995) “Practical Coordination Control between Satellite Attitude and Manipulator Reaction Dynamics Based on Computed Momentum Concept.” IROS ‘95.
  • Clohessy W.H., Wiltshire R.S. (1960) “Terminal Guidance System for Satellite Rendezvous.” J Aerospace Sciences 27.
  • Bekker M.G. (1969) Introduction to Terrain-Vehicle Systems. Univ Michigan Press.
  • Wong J.Y. (1989) Terramechanics and Off-Road Vehicles. Elsevier.
  • Stentz A. (1994) “Optimal and Efficient Path Planning for Partially-Known Environments.” ICRA ‘94 (Field D*).
  • Bualat M., Smith T., Smith E., Fong T., Wheeler D. (2018) “Astrobee: A New Tool for ISS Operations.” AIAA SpaceOps Conference.
  • Maimone M., Cheng Y., Matthies L. (2007) “Two Years of Visual Odometry on the Mars Exploration Rovers.” J Field Robotics 24(3).
  • Verma V., Maimone M., Gaines D., Francis R., et al. (2023) “Autonomous Robotics is Driving Perseverance Rover’s Progress on Mars.” Science Robotics 8(80).
  • NASA-STD-8719.13C — Software Safety Standard.
  • ECSS-Q-ST-60-15C — Radiation hardness assurance for EEE components.
  • NASA-HDBK-4002A — Mitigating In-Space Charging Effects.

Graph View