FIELD STUDY / 06
Space robotics
New missions beyond Earth
Will a movement learned on Earth work in space?
UNDERSTAND THE FIELD
Beyond the name of a technology, look at the work it changes.
Editorial selection · source snapshot 23 Oct 2026
THE BIG PICTURE
The big picture
Beyond Earth, familiar manipulation problems become new questions. Capturing an unsupported object requires reasoning about motion and contact together. We read research to distinguish current abilities from challenges that remain.
THE OPEN PROBLEM
The challenge ahead
Capturing freely moving objects in an unfamiliar environment. Results from a narrowly trained policy and a generalist must not be mistaken for a controlled, like-for-like comparison.
IN THE WORLD
Where the work leads
Begin with benchmarks for capturing objects inside spacecraft and manipulating them in space-related environments.
BEYOND THE OVERVIEW
A closer look at the work
Choosing the exploration goal and reading the ground
A rover choosing a local path does not necessarily choose the scientific purpose of the mission. Which terrain to examine and which measurements matter belong to the team's science plan. The shape of a nearby rock or a narrow passage, meanwhile, appears in observations gathered close to the ground. Dividing these decisions connects human planning with local autonomy. This differs from quickly teleoperating a robot inside a building. Approving every small route change from Earth would insert a communication wait between actions. Defining which decisions the rover makes locally and which situations it reports is therefore part of mission design.
Exploration continues when the data comes home
A traverse informs the next plan on Earth only when images, location information and equipment status return. NASA's account of Mars communications describes orbiters relaying much of the surface missions' science data to Earth. Movement, observation and transmission are therefore connected activities. What is collected and when it is delivered shape the information available for the next decision. An interesting terrain feature, for example, is easier to interpret when its location and surrounding context are recorded with it. This is a perspective on exploration, not a reconstruction of a particular mission's transmission schedule. A robot's contribution includes the observations and next questions produced by its travel, as well as the distance covered.
WHAT TO LOOK FOR
What to notice in the next story
Separate capture results for static and moving targets.
Check how generalist and object-specific policies were trained.
AstroBench's 2/200 and 8/10 come from separate experiments.
READ THE ORIGINAL
Sources behind this perspective
Open resources selected as a starting point for this field. The originals describe the experimental conditions and scope of the work.