A 2-DoF on-line Template Matching Strategy for LIDAR-based Pose Initialization of Non-Cooperative Spacecraft

On-Orbit Servicing and Active Debris Removal missions require close-proximity operations with resident space objects. In such scenarios, accurately estimating the pose of non-cooperative targets represents a major challenge due to the absence of markers and prior state information. To this end, an innovative model-based pose initialization architecture based on LIDAR measurements is proposed. Pose estimation is first carried out using an on-line Template Matching strategy in which the pose search space is reduced to 2 Degrees of Freedom exploiting centroiding and Principal Component Analysis. The achieved solution is then refined by applying the Iterative Closest Point algorithm. The performance is numerically assessed using an in-house LIDAR simulator to realistically replicate the point cloud generation process, taking as target the ENVISAT satellite. Results demonstrate that the technique achieves a success rate exceeding $60 \%$ across a broad range of target distances and orientations, peaking at $94 \%$ at 50 meters, with an average attitude error of 0.972° and positional accuracy of 0.234 m. The study further compares two weighting schemes for point-to-point association in the pose refinement step, showing superior performance with exponential weighting.