LiDAR Attitude Estimation Based on LiDAR Point-Cloud Data Processing

Light detection and ranging (LiDAR) sensors are becoming increasingly popular because of their ability to provide a fine 3-D representation of the surrounding scene. Such a representation is precious for several applications, like simultaneous localization and mapping (SLAM), road monitoring, and autonomous or semi-autonomous driving, to cite a few. In order to profitably use the data, however, it is required to know the orientation of the LiDAR, referred to as attitude. Commonly, the estimation of the attitude is made possible by inertial measurement units (IMUs), which are anchored to the LiDAR and measure both the angular velocity and acceleration. In fact, the attitude estimation can be performed by means of different approaches that integrate the angular velocity and/or single out from the acceleration the gravitational contribution, which informs about the attitude. The achievable accuracy fundamentally depends on the uncertainty of the input data, that can be affected by offset, drift, and noise. A method is proposed to improve the attitude estimation by exploiting the information contained in the LiDAR output, namely the point-cloud that represents the surrounding scene. Specifically, the point-cloud and IMU data are processed by means of a Kalman filter and a strapdown integration strategy. The proposed method is validated both through simulations and experiments.