23–24 Oct 2026
Kraków, Poland
Europe/Warsaw timezone

Machine Learning Detection of Penetration Anomalies in Planetary Lander Data: A Cross-Mission Validation Using MUPUS, HP3, and SSP

23 Oct 2026, 12:00
1h
Uniwersytet Jagielloński, Wydział Fizyki, Astronomii i Informatyki Stosowanej (Kraków, Poland)

Uniwersytet Jagielloński, Wydział Fizyki, Astronomii i Informatyki Stosowanej

Kraków, Poland

ul. prof. Stanisława Łojasiewicza 11
Board: 10

Speaker

Poojitha Annabathula

Description

Planetary lander penetrometer and subsurface probe experiments have repeatedly encountered unexpected mechanical resistance, including MUPUS on Rosetta's Philae lander (comet 67P), the HP3 "mole" on NASA's InSight lander (Mars), and the Surface Science Package on ESA's Huygens probe (Titan). Each anomaly was identified and characterized manually by mission teams, with distinct underlying causes ranging from a sintered dust-ice crust to insufficient regolith friction.

This study develops a change-point and pattern-deviation detection method, calibrated on the MUPUS penetrometer dataset from the Philae lander's First Science Sequence, where the instrument's 216-cycle hammering sequence failed to achieve expected depth despite escalating energy levels. MUPUS penetrometer data, developed by the Space Research Centre of the Polish Academy of Sciences, serves as the primary calibration case, given its independently confirmed anomaly (Spohn et al. 2015).

The method is then applied, without modification, to publicly archived HP3 and SSP datasets to test whether it correctly and independently identifies known anomalies of different physical origin and signature shape across missions and planetary bodies. We report detection performance across all three datasets, discuss the method's ability to distinguish anomaly types, and consider its potential as an automated screening tool for future subsurface probe missions, including ESA's upcoming Comet Interceptor.

Author

Presentation materials

There are no materials yet.