TL;DR

Researchers from the UK and Japan have developed a machine learning algorithm called YOLO-ETA to locate the landing site of Luna 9, the first human-made object to land safely on the Moon in 1966. The algorithm was trained on Apollo landing site imagery and has already identified several promising candidate locations.

A Six-Decade-Old Mystery

Luna 9 touched down on the Moon on 3 February 1966, marking a milestone in space exploration. But the coordinates published in Pravda at the time carried high uncertainty, potentially placing the actual site tens of kilometres from the reported position. Decades later, the precise landing location remains unconfirmed.

Lewis Pinault at University College London led the research, published in npj Space Exploration. The team developed YOLO-ETA — short for “You-Only-Look-Once — Extraterrestrial Artifact” — a machine learning algorithm trained to recognise the visual signatures of spacecraft landing sites in lunar orbital imagery captured by NASA’s Lunar Reconnaissance Orbiter Camera (LROC).

Training on Known Sites

The algorithm learned what landing sites look like by studying imagery of Apollo mission locations, where the precise coordinates are well established. The team then validated the approach by successfully identifying the known landing site of Luna 16, a Soviet sample-return mission from 1970.

With that confirmation in hand, the researchers applied YOLO-ETA to a 5-by-5 kilometre region around the published Luna 9 coordinates. The algorithm flagged several promising candidate sites that match the expected visual characteristics of a 1960s-era spacecraft landing.

Looking Forward

Confirmation may come sooner than expected. India’s Chandrayaan-2 orbiter is scheduled to pass over the region in March 2026, and its high-resolution imaging could provide the evidence needed to verify or rule out the candidates identified by YOLO-ETA. If successful, the technique could be applied to other lost or imprecisely located spacecraft from the early space age, turning orbital imagery archives into searchable databases for space archaeology.