Unsupervised Remittance Calibration for
Multi-Beam Lidar · 2010
In recent years, multi-beam Light Detection and Ranging (LIDAR) sensors have become increasingly significant in autonomous navigation. While use of multi-beam sensors does have multiple benefits, such as increased resolution, it also significantly complicates calibration, including calibration of remittance returns. Consistent remittance returns from LIDAR sensors are particularly valuable for autonomous perception and action in real-world scenarios. Our goal is to develop an unsupervised remittance calibration approach which leverages the exceptional amount of data produced by the LIDAR sensor and the platform vehicle's change in pose to bring the multiple individual-beam sensors into agreement. We address this problem with 2 approaches: a simple conditional expected value over multiple beams 1 and a modified Expectation-Maximization (EM) algorithm 2 . Our calibration shows a decrease in variance between the remittance observances of individual beams while still maintaining the dynamic range of possible sensor outputs.