Complete prepared-input INS-LiDAR evaluation reuse

This archive retains all six complete fresh-process runs, versioned prepared problem, generator, original public OXTS bytes, full source capsules/manifests, measurement code/schema, public result schema, complete residual/optimizer histories, counters and validation receipts. index.json binds every publication byte other than itself.

The reference uses all 2,762 OXTS poses of KITTI 2011_09_30_drive_0018_sync. Sensor poses 10-2751 and their 788 relative motions are generated using the declared extrinsic, clock, scale and seeded noise. No raw LiDAR registration is performed. This is a performance/numerical-parity control, not independent calibration accuracy or raw-pipeline evidence.

KITTI attribution: Andreas Geiger, Philip Lenz, Christoph Stiller and Raquel Urtasun. Original data and its derived control retain CC BY-NC-SA 3.0: https://www.cvlibs.net/datasets/kitti/ and https://creativecommons.org/licenses/by-nc-sa/3.0/. Calibrex source snapshots and measurement code retain Apache-2.0.

raw-oxts.tar.gz contains all original OXTS packet/timestamp bytes and the original download provenance. The audit reconstructs every reference pose and generated sensor motion. history.json.gz retains every normalized evaluation's parameters/residual digest and every least-squares result's parameters, termination counters and complete residual/Jacobian digests. Real full and short Git identities are verified separately before comparison.

interrupted-v1 retains an earlier incomplete cohort and its audit correction; none of its timing is included in the final six-run comparison.

Audit from the repository root without calibration or raw downloads: python -m tools.benchmark_trajectory_hand_eye_reuse --audit-only docs/assets/trajectory_hand_eye_evaluation_reuse