FURI | Spring 2026

Event-Driven Spacecraft Pose Estimation via CNN Direct Regression from Voxel Grid

Data icon, disabled. Four grey bars arranged like a vertical bar chart.

6 degrees of freedom (6-DoF) spacecraft pose estimation is essential for autonomous orbital operations and is still a challenge under extreme lighting conditions. This research introduces a deep learning pipeline that processes asynchronous data from event cameras into exponential-decay voxel grids for direct pose regression. By comparing three architectural variants, the team identifies the synthetic-to-real domain gap as the primary performance bottleneck for their architecture. These findings provide critical insights for optimizing neuromorphic vision systems in real-time space robotics and navigation.