3D Lidar-IMU Calibration Based on Upsampled Preintegrated Measurements for Motion Distortion Correction

Abstract

In this paper, we present a probabilistic framework to recover the extrinsic calibration parameters of a lidar-IMU sensing system. Unlike global-shutter cameras, lidars do not take single snapshots of the environment. Instead, lidars collect a succession of 3D-points generally grouped in scans. If these points are assumed to be expressed in a common frame, this becomes an issue when the sensor moves rapidly in the environment causing motion distortion. The fundamental idea of our proposed framework is to use preintegration over interpolated inertial measurements to characterise the motion distortion in each lidar scan. Moreover, by using a set of planes as a calibration target, the proposed method makes use of lidar point-to-plane distances to jointly calibrate and localise the system using on-manifold optimisation. The calibration does not rely on a predefined target as arbitrary planes are detected and modelled in the first lidar scan. Simulated and real data are used to show the effectiveness of the proposed method.

IN2LAMA: INertial Lidar Localisation And MApping

“…We can see that the accuracy of the estimate increase with the length of the dataset. These results demonstrate similar accuracy as our previous work [2] that relies on observing a calibration target made of at least three planes.…”

“…By doing so, the inertial integrations can be computed beforehand and the resulting pseudo-measurements stay constant during the estimation process, regardless of the linearisation point changes. In our earlier work on lidar-IMU extrinsic calibration [2], we extended this concept by considering preintegration over continuous representation of the inertial data. The UPMs allow the computation of inertial data at any timestamp and therefore, enable the characterisation of the system's motion during a lidar scan without the use of any explicit motion/statemodel.…”

“…The proposed framework uses UPMs to accurately address the problem of motion distortion in lidar scans. We previously introduced these measurements in [2] based on concepts originally presented in [3] and [4]. The original idea of preintegration [3] consists in reducing the size of the state to estimate by combining IMU measurements between two estimated poses.…”

“…is used as a plane feature residual. As in [2], the variance of lidar residuals requires the knowledge of the state. Therefore, the noise propagation needs to be executed regularly during the optimisation.…”

“…In our previous work [2], we have shown that the global weight of the lidar factors in the optimisation depends on the arbitrarily set number of features used per frame. This chosen number does not necessarily represent the correct confidence of the lidar registration.…”