Abstract
The multi-modal perception methods are thriving in the autonomous drivingfield due to their better usage of complementary data from different sensors.Such methods depend on calibration and synchronization between sensors to getaccurate environmental information. There have already been studies aboutspace-alignment robustness in autonomous driving object detection process,however, the research for time-alignment is relatively few. As in realityexperiments, LiDAR point clouds are more challenging for real-time datatransfer, our study used historical frames of LiDAR to better align featureswhen the LiDAR data lags exist. We designed a Timealign module to predict andcombine LiDAR features with observation to tackle such time misalignment basedon SOTA GraphBEV framework.