Multi-sensor calibration for an unmanned delivery robot

Abstract

The last mile delivery industry is a huge and ever-growing industry with increasing demand for manpower needs. Developments into the use of autonomous mobile robots to address the industry’s manpower demand requirements have been underway. The NTU-Continental Lab have been developing their own autonomous mobile robot named Corriere to facilitate the research and development for this purpose.

The purpose of this project involves the implementation and analysis of calibration algorithms for the different sensors on-board the Corriere robot for the sake of accurate sensor fusion. In this report, we cover the successful implementation of various offline calibration tools to calibrate three types of sensors for the Corriere delivery robot. In total, three types of sensors are discussed: 3D LiDAR, RGB camera and IMU. This project covers four parts. The first two parts involve the intrinsic calibration of the IMU and camera was implemented using the allan_variance_ros and Kalibr calibration tools to obtain the intrinsic calibration parameters. The second two parts involve using these intrinsic calibration parameters were used for the extrinsic calibration between the LiDAR – camera using LVT2Calib and IMU – camera using Kalibr. The report first covers the method of data collection and implementation of the specific calibration tools. Subsequently, the best calibration results obtained empirically from each calibration tool is analyzed and discussed.