AI Improves Parachute Landing Accuracy

ycombinator · x · 2026-07-14

[Repost] Inversion Space explains how they use the latest computer vision and AI models to track the movement and deformation of a parachute canopy under various flight conditions.

Using data from over 100 flights, covering different control inputs, visual disturbances, canopy deflections, and attitude changes, they fed these observations into an 8-DOF model to further tune Arc's GNC algorithms. The result was a 90% reduction in landing error. Their goal is to make flexible canopy wings nearly as controllable as rigid metal wings.

Original post →

More from Embodied

Embodied channel →