A robust online extrinsic calibration method for GNSS-RTK and IMU system and vehicle setups

Peng Wang, Qingyang Xu, Siyun Yao, Xiangfei Wu, Qunli Zhang, Ruichen Lv, Shiming Liu, Yamin Li, Hui Fang, Zheng Hu

Abstract: To guarantee safe operation of autonomous vehicles, multi-modal sensors such as cameras/lidars/ integrated navigation systems (INS) are utilized to perceive the environment and ego's location, pose and kinematic information. However, the perception and localization results are inaccurate or even incorrect when spatial relationship between sensor coordinates and vehicle coordinates is not built correctly. That further leads to high safety of the intended functionality (SOTIF) risks (degradation or even failure of the functionality) and causes safety threats. This work focuses on accurate and trustworthy estimation of the misalignment angle between INS coordinates and vehicle coordinates even with low-accuracy/cost equipment under harsh conditions. The proposed method achieves highly accurate and consistent results in a bumpy and signal-instable opencast environment. To enhance the dependability of the method, verification algorithms are designed to monitor the calibration quality, infer the potential failure reason, and provide proper user guidance accordingly.