SLAC

In Calibrex, SLAC means simultaneous localization and calibration as an architecture principle rather than one fixed algorithm.

The core representation must be able to express trajectory, extrinsics, timing, intrinsics, map variables, correspondences, priors, and evaluation diagnostics.

Calibrex should not hard-code one SLAC algorithm into the core. Instead, it compiles a calibration problem from config into backend-neutral variables, factors, priors, gauges, and evaluation requirements. Sensor-specific methods then enter as plugins, optional solver adapters, or baseline subprocesses.

Research Anchors

Autonomous-driving and fixed-rig LiDAR calibration work is guided by these papers:

• SceneCalib: targetless multi-camera and LiDAR self-calibration for autonomous driving. It motivates joint camera intrinsics, camera-camera extrinsics, and camera-LiDAR extrinsics.
• SST-Calib: simultaneous spatial-temporal LiDAR-camera calibration. It makes time offset a first-class calibration variable rather than metadata.
• Decentralized multi-LiDAR SLAC: joint vehicle pose, velocity, map, and multi-LiDAR extrinsic estimation. It is the cleanest reference for LiDAR-centered SLAC.
• M-LOAM: online multi-LiDAR odometry, mapping, and extrinsic calibration. It informs a future Calibrex online design, but not the first offline MVP.
• Koide et al. ICRA 2023: general, single-shot, targetless, automatic LiDAR-camera calibration. Calibrex should support it as an adapter or baseline before trying to duplicate it natively.

See LiDAR SLAC for the concrete design consequences.