Embodied Synthetic EVA
Research program / Human spaceflight / Synthetic embodiment

Embodied
Synthetic EVA

A shared-agency architecture for projecting human sensorimotor presence into synthetic bodies — keeping biological crew protected while human judgment remains physically present where it matters.

01 Tele-Embodiment 02 Shared Agency 03 Machine Somatosensation 04 Synthetic Physiology 05 Earth → Moon → Mars
Read the research ↓ Development path
Core propositionThe machine goes.
The human can still be there.
Human roleJudgment + improvisation
Machine roleSurvivability + execution
Near-term arenaHazardous terrestrial work
DestinationMoon / Mars surface ops
01 / Abstract

Human capability, projected beyond biology.

A concept architecture for letting astronauts perform physical work through synthetic bodies while autonomy handles stabilization, reflexes and increasingly large fractions of execution.

AbstractConcept Architecture

Future human exploration of the Moon and Mars will require physical work in environments that impose substantial physiological, operational and logistical burdens on biological crew. Conventional extravehicular activity places astronauts directly within vacuum, radiation, abrasive dust, thermal extremes and equipment hazards, while autonomous robotic systems remain poorly suited to many unstructured manipulation and contingency tasks requiring human judgment.

Embodied Synthetic EVA proposes an intermediate architecture in which an astronaut projects sensorimotor agency through a remotely located robotic or synthetic body while machine autonomy provides stabilization, reflexes, task execution and increasing levels of independent operation.

The architecture extends historical work in telepresence, robotic EVA, Surrogate Astronaut Robotic Avatars, NASA surface telerobotics and ESA/DLR Surface Avatar by treating embodiment itself as an engineering variable. Control may transition continuously between direct human tele-embodiment, shared control, supervised autonomy and autonomous operation.

tele-embodimentshared autonomyhuman spaceflightsynthetic physiologyhapticsmachine somatosensationlunar operations
Paper at a glance15 sections
E0→E6
Embodiment continuum
6
Architecture layers
4+
HSF mission modes
4
Earth precursor sectors
Central proposition

The machine goes, but the human can be there when needed.

02 / Core Thesis

Decouple human presence from the human body.

Spaceflight has historically treated biological presence and operational presence as the same thing. They do not need to be.

Future sustained lunar and Martian operations make the traditional choice between “send the astronaut” and “send the robot” increasingly restrictive. Human judgment, improvisation and dexterity remain exceptional, but biological exposure to vacuum, radiation, abrasive regolith, thermal extremes and complex EVA operations is costly.

The more useful question is therefore not whether a task belongs to a human or a robot. It is:

How can human perception, judgment, dexterity and improvisation be projected into a hazardous environment without requiring the biological human body to occupy that environment?

Embodied Synthetic EVA treats the surrogate as an alternative operational body. The astronaut may directly inhabit it through synchronized sensorimotor interfaces, while local intelligence manages balance, contact, navigation, reflexes and routine action. The long-term objective is neither full teleoperation nor full autonomy. It is transferable physical agency.

03 / Research Lineage

Not a new robot. A new relationship with the robot.

The novelty is not the existence of a robotic surrogate astronaut. The contribution is the integration of embodiment, shared agency, somatosensation and synthetic physiology into one measurable HSF architecture.

1980s–1990s
Remote work for lunar bases
NASA studies explored remotely operated equipment for hazardous, repetitive and time-consuming lunar-base operations.
1998
Human–robot cooperative EVA roles
Akin formalized roles ranging from robotic assistant to surrogate, specialist and human–robot symbiosis.
2013
SARA — Surrogate Astronaut Robotic Avatar
Chau and Thangavelu explicitly proposed humanoid robotic avatars performing work in place of exposed astronauts.
2010s–2020s
NASA surface telerobotics + HERRO
ISS-to-surface operations and Mars-orbit telerobotics established that cognitive presence and physical location can be separated.
2023–2026
METERON / Surface Avatar
ESA/DLR demonstrated multimodal interfaces, haptic force reflection, direct teleoperation, task delegation and intervention during autonomous execution.
Proposed next step
Embodiment as the system requirement
Preserve a coherent body model while agency transfers between human and machine, and extend the surrogate toward synthetic somatosensation and physiology.
04 / Embodiment Continuum

Agency is a slider, not a switch.

A mission may move repeatedly between direct human control and machine autonomy. The body stays the same; the locus of agency changes.

E0
Biological EVA
Human: physical actor
Machine: tool/support
E1
Remote Operation
Human: command
Machine: low-level execution
E2
Immersive Telepresence
Human: direct operator
Machine: stabilization
E3
Tele-Embodiment
Human: embodied actor
Machine: reflex + support
E4
Shared Embodiment
Human: intent + selective control
Machine: task execution
E5
Supervisory Surrogate
Human: goals + intervention
Machine: predominant execution
E6
Autonomous Synthetic Agent
Human: optional
Machine: autonomous
The objective is not E6. The objective is fluid, safe transfer of embodiment in both directions.
Novel event

Human assumes embodiment

Direct first-person control resolves an unfamiliar connector, damaged mechanism, science target or contingency.

⇄
Agency handoff
Routine execution

Machine resumes autonomy

Local AI handles locomotion, balance, transport, inspection and repetitive actions until human judgment is again required.

05 / System Architecture

Six layers from human intent to synthetic body.

The architecture preserves sensorimotor coherence while allowing morphology, sensing and intelligence to depart from ordinary human anatomy.

01
Input + Return

Human Sensorimotor Interface

Capture intention and return external sensory information without requiring every channel to be reproduced literally.

  • motion
  • gaze
  • EMG
  • voice
  • haptics
  • vision
02
Translation

Embodiment Mapping

Retarget movement, scale force and reconcile different body geometries while preserving predictable sensorimotor relationships.

  • retargeting
  • force scaling
  • morphology mapping
03
Control

Shared Agency

Human operates at the right abstraction level while the machine handles balance, grip regulation, collision avoidance and local planning.

  • intent
  • stabilization
  • task skills
  • handoff
04
Body Sense

Machine Somatosensation

Distributed touch, strain, slip, temperature, loading and damage sensing create a synthetic peripheral nervous system.

  • tactile skin
  • strain
  • slip
  • reflexes
05
Mechanics

Synthetic Physiology

Progress from conventional motors toward compliant actuation, tendon systems, artificial muscles and adaptive structures.

  • compliance
  • tendon
  • muscle
  • thermal paths
06
Evolution

Environment-Adapted Morphology

Retain enough anthropomorphism for embodiment while changing anatomy where vacuum, dust, terrain and tool use reward it.

  • IR vision
  • tool coupler
  • extra limb
  • replaceable skin
The human form becomes a control reference — not a design prison.
06 / Spaceflight ConOps

From pre-EVA scouting to persistent synthetic crew assets.

The architecture supports several operating modes without requiring a fully autonomous humanoid.

ConOps A

Pre-EVA

The surrogate goes first so biological EVA time is spent only where direct human presence adds value.

  • Inspect route and terrain
  • Stage tools and consumables
  • Clean and prepare interfaces
  • Verify power, comms and vehicles
ConOps B

Surrogate EVA

The astronaut remains inside a protected, shirtsleeve environment while physically acting through the surrogate.

  • Exterior maintenance
  • Cable and connector work
  • Instrument installation
  • ISRU and infrastructure servicing
ConOps C

Collaborative EVA

A biological astronaut outside works with one or more surrogates operated by crew or specialists inside.

  • Second pair of hands
  • Heavy-lift support
  • Remote specialist injection
  • Safety / rescue asset
ConOps D

Orbit-to-Surface Embodiment

Low-latency orbital crews project physical agency onto a planetary surface without landing the biological operator.

  • Lunar orbit → Moon
  • Mars orbit → Mars
  • Science + maintenance
  • Reduced surface exposure
07 / Terrestrial Precursors

Space qualification should not be the first time it meets real work.

Earth provides commercially meaningful environments that mature the same sensing, control and embodiment stack needed for the Moon and Mars.

N

Nuclear

Remote maintenance, inspection, waste handling and decommissioning where skilled manipulation is valuable but radiation exposure is not.

M

Mining

Dust, unstable terrain, heavy machinery, poor visibility and restricted communications create a direct analogue for lunar industrial operations.

D

Disaster Response

Unknown environments force the human/autonomy handoff hypothesis: machine handles routine work, human enters when novelty exceeds capability.

R

Remote Infrastructure

Offshore energy, tunnels, polar facilities, hazardous chemical sites and extreme construction create early deployment pathways.

These are not merely spin-offs. They are upstream technology maturation environments for human spaceflight.
08 / Experimental Framework

Make embodiment falsifiable.

The program should be judged by operational performance and human factors — not whether the demo looks futuristic.

An initial Earth testbed can use a commercially available mobile manipulator or humanoid, immersive display, motion tracking, haptic feedback, instrumented tasks and a shared-autonomy layer. The same tasks should be repeated under conventional remote control, immersive telepresence, tele-embodiment, shared embodiment and supervised autonomy.

HIB = Thuman control / TmissionHuman Intervention Burden — one proposed core measure
Operational
  • task success
  • completion time
  • errors + collisions
  • recovery rate
  • energy use
Human
  • NASA-TLX
  • agency
  • body ownership
  • situational awareness
  • fatigue
Interface
  • visual latency
  • haptic latency
  • mapping error
  • sensor bandwidth
  • dropped data
Mission
  • EVA hours displaced
  • crew-hours recovered
  • exposure avoided
  • consumables saved
  • availability gained

Research hypotheses: embodiment should improve complex manipulation; shared agency should outperform pure teleoperation or pure autonomy in mixed-predictability tasks; operators should tolerate bounded morphological divergence; local machine reflexes should reduce bandwidth and intervention burden; and surrogate architectures should reduce mission-level biological exposure.

09 / Synthetic Physiology

From motors and gearboxes toward engineered anatomy.

Synthetic physiology is a long-term evolutionary layer, not a prerequisite for first deployment.

Near term

Electromechanical body

Commercial humanoid or mobile manipulator. Mature actuation, teleoperation and sensing. Build the embodiment stack first.

→
Longer term

Synthetic physiology

Compliant mechanisms, tendons, artificial muscles, distributed skin, local reflexes, adaptive structure and selected biohybrid components.

The attraction of artificial muscle and machine skin is not cosmetic realism. It is compliant interaction, compact distributed actuation, shock tolerance, richer body sensing and a mechanical architecture that may map more naturally onto human sensorimotor control.

Biohybrid robotics pushes further by integrating living tissue with engineered systems. For spaceflight it remains a long-horizon research direction because of radiation, vacuum, nutrition, sterility, reliability and planetary-protection constraints. Its relevance here is architectural: the system should not assume the surrogate will always be a rigid motorized robot.

10 / Morphology

How non-human can a human learn to inhabit?

The surrogate should preserve enough anthropomorphism for intuitive embodiment while exploiting the freedom of an engineered body.

Human-built environments reward human-compatible reach, tools and work envelopes. But the Moon does not reward strict anatomical fidelity. A surrogate might lower its center of gravity, add a manipulator, rotate joints beyond human limits, swap hands for tools, integrate anchors, widen its spectral vision or treat structural damage as a replaceable module rather than an injury.

The long-term research program may be less about reproducing human anatomy — and more about discovering how much extra capability a human nervous system can learn to inhabit.
11 / Embodiment Geography

Latency determines how deeply a body can be shared.

Embodiment should be treated as a capability whose achievable depth changes with communications geometry.

H

Habitat → Surface

Very low latency. Strong candidate for high-fidelity haptic embodiment and direct whole-body control.

L

Lunar Orbit → Moon

Low enough latency for interactive telerobotics; strong candidate for tele-embodiment with local machine reflexes.

E

Earth → Moon

Delay reduces transparency. Shared autonomy and task-level execution become more important.

M

Mars Orbit → Mars

Potentially powerful architecture for real-time surface exploration without immediate crew surface exposure.

12 / Development Roadmap

Start with one limb. End with persistent synthetic crew assets.

A staged program can create useful evidence long before a full synthetic astronaut exists.

Phase I

Embodied Limb

Arm/hand mapping, stereo vision, force feedback, tool and connector tasks.

Phase II

Mobile Surrogate

Whole-body operation, locomotion, manipulation, balance and human/AI transition.

Phase III

Hazardous Earth Demo

Nuclear mockup, mining site, industrial maintenance or disaster training environment.

Phase IV

Synthetic EVA Analogue

Compare suited simulation, remote operation, embodied control and shared autonomy.

Phase V

Space Demonstration

ISS/Gateway crew controlling a remote embodiment platform under flight constraints.

Phase VI

Persistent Synthetic Crew

Infrastructure assets work autonomously before arrival, become embodied during crew presence, then continue afterward.

Laboratory embodiment → hazardous terrestrial work → lunar analogue → crew-controlled remote operations → Synthetic EVA → persistent Moon/Mars surrogate systems.
13 / Safety + Ethics

The control handoff is itself a safety-critical system.

Embodiment creates new failure modes: agency confusion, automation surprise, sensory conflict, cybersickness and over-trust.

A formal Agency Authority Model should always define the current controller, permitted control envelope, machine safety vetoes, transfer request, transfer acknowledgement, emergency return and communications-loss behavior.

The architecture does not require machine consciousness, artificial personhood or human-equivalent cognition. Early surrogate systems are best treated as instruments of extended human action. Ethical questions around increasingly human-like or biohybrid machines may emerge later, but they should not be conflated with the near-term engineering requirement.

14 / Conclusion

The astronaut may not always step outside.

Future exploration can evolve from transporting the biological human body everywhere capability is required toward transporting human agency.

Embodied Synthetic EVA reframes human–robot collaboration around a continuous relationship between biological operator and synthetic body. Human intelligence supplies improvisation, judgment, semantic understanding and expertise. The machine supplies survivability, strength, endurance, sensing, reflexes and autonomous routine execution.

As AI improves, direct human involvement can decline without invalidating the architecture. At first: human intelligence → machine body. Later: human intention → machine execution. Eventually: human intervention → only when desired or necessary.

The astronaut of the future may not always step outside the habitat. Sometimes, the astronaut may simply change bodies.
15 / References

Technical lineage and supporting literature.

Selected references from the Springer manuscript. Links open to the original source where available.

NASA. Uncrewed Lunar Surface Operations and Support Activities, 2022. NASA NTRS ↗
NASA. Evidence Report: Risk of Injury and Compromised Performance due to EVA Operations, 2022. NASA NTRS ↗
NASA. EVA Development in Exploration-Driven Human Spaceflight, 2025. NASA NTRS ↗
Akin, D. L. “Experimental Investigation of Cooperative Human/Robotic Roles in Extravehicular Operations,” 1998.
Chau, A. T., Thangavelu, M. “Surrogate Astronaut Robotic Avatars: Co-Robotics for Safe, Economic Space Operations,” AIAA SPACE 2013.
Fong, T. et al. Testing Astronaut-Controlled Telerobotic Operation of Rovers from the ISS as a Precursor to Lunar Missions. NASA NTRS ↗
NASA. HERRO: A Science-Oriented Strategy for Crewed Missions Beyond LEO. NASA NTRS ↗
DLR. Surface Avatar — Multimodal Teleoperation and Scalable Autonomy. DLR ↗
DLR. “Enhancing Scalable Autonomy Space Teleoperation with User Intervention During Task Execution,” IEEE Aerospace, 2026.
Toet, A. et al. “Toward Enhanced Teleoperation Through Embodiment,” Frontiers in Robotics and AI, 2020. Frontiers ↗
“Towards Human-Like Tactile Perception in Humanoid Robot Forearms,” Biomimetic Intelligence and Robotics, 2026.
“A Tactile Reflex Arc for Physical Human-Robot Interaction,” Mechatronics, 2025.
Afsar, O. K. et al. “Electrofluidic Fiber Muscles,” Science Robotics, 2026.
Garmroudi, A. et al. “Biohybrid Living Robotics,” npj Robotics, 2025.
International Atomic Energy Agency. Guidance on remote techniques for minimizing personnel exposure during nuclear operations and decommissioning.
NIOSH. Mining automation and safety research addressing robotics, teleoperation, sensing, communications and human factors.
NIST. Response robot evaluation and standardized emergency-response test methods.