1. The pod’s eye view

In 2001: A Space Odyssey, the EVA pod approaches the monolith with agonizing slowness. The tension comes from the mismatch between human perception and the precision required. Dave does not have a laser scanner; he has windows, practice, and a lot of patience. Real proximity operations are moving toward doing the same thing without the windows.

This entry reads about the sensors that make autonomous close approach possible.

2. The sensor stack

The literature describes a hierarchy of sensors, each useful at a different range:

  • GPS and carrier-phase differential GPS: works at tens of kilometers down to a few meters, provided both spacecraft have good sky view and the geometry is favorable. It is the cheapest relative-navigation layer but fails close in, where multipath and signal blockage matter.
  • Radio ranging and inter-satellite links: time-of-flight measurements over UHF or S-band can give range and range-rate at medium distances. They are robust but provide only one dimension directly; angles must be inferred or combined with other sensors.
  • Vision: cameras with natural-feature tracking or fiducial markers are the workhorse for final approach. They give angle and, with stereo or known target size, range. They are lightweight, low power, and well understood. Their weakness is lighting: a target in eclipse, or sun glint, can confuse algorithms.
  • Lidar and laser triangulation: provide accurate 3-D shape at close range and do not depend on illumination. The cost is mass, power, and complexity; for small satellites this is still a premium choice.
  • Radar: useful when the target is non-cooperative or unlit, but miniaturized space radar is heavier and hungrier than vision.

3. What the practitioners say

The ScienceDirect overview notes that vision-based methods using artificial markers have been extensively studied but are sensitive to lighting and target attitude. The CETACEAN project at NASA is essentially an attempt to modularize the software side: take a set of sensor inputs, fuse them onboard, and output a reliable relative state for proximity operations and docking. Starfish Space’s 2025 LEO demonstration with Impulse Space is the kind of flight heritage the literature keeps pointing at — autonomous rendezvous and close-proximity maneuvers between two small spacecraft, not merely a chase and pass.

The common thread is redundancy and range. No single sensor covers the whole approach from kilometers to contact. A practical system layers them and switches or fuses as the range closes.

4. What this means for the desktop

For a rack-based desktop, almost none of this is needed. The cells are mechanically registered; relative navigation is a mechanical problem, not a sensor problem.

For a free-flying desktop, it is table stakes. Every cell needs enough sensing to find its neighbors and enough compute to fuse the data. That is a different cost model than a rail and a blind-mate connector.

5. What this changes

  • Relative navigation is added to the list of things that make free-flying cells expensive.
  • Vision-based final approach is identified as the most plausible near-term sensor layer if docking is ever required.
  • The rack’s mechanical registration is reframed as a navigation simplification: it removes a whole sensor fusion problem.