I am working on a standalone simulation and I found that the camera images (including depth images and point clouds) are always delayed 2 rendering frames. For example, if rendering frequency is 10Hz, the rendered image is 0.2s delayed image. If the camera is moving in 1m/s, the image captured from 0.2m behind the actual camera position is obtained. Since the images are delayed, timestamps of them are not matched with their actual rendered time. This problem is making inaccurate and biased results when I use the simulation for visual odometry or navigation. How can I resolve this problem?
The posts in the link below seems to be talking about the result of the same issue.