How do you approach the calibration of autonomous vehicle sensors to ensure data accuracy and reliability?
Instruction: Discuss the processes and techniques for calibrating different types of sensors used in autonomous vehicles.
Updated
Example Answer
Calibration establishes how measurements relate to reference quantities and coordinate frames. I would estimate camera intrinsics and distortion, inter-sensor extrinsic transforms and relevant timing offsets, then verify them against independent observations. A Kalman filter can fuse measurements under a model; it does not automatically calibrate every geometric discrepancy. Monitor drift and alignment residuals, but only accept recalibration changes when the available data and validation criteria support them.
Make it your own
Illustrative diagnostic: plot projected LiDAR points over images for a static target and a moving target. If only the moving case misaligns, investigate timing as well as geometry.
Why this works
Different residual patterns suggest different causes. Follow-up: how would you prevent an unobservable online calibration update from degrading a correct transform?
Sources: OpenCV camera calibration, OpenCV calibration and reconstruction.