Running Kimera-VIO in Gazebo or on simulated data?

[shonigmann]

Since I noticed that one of the sample datasets for Kimera-Semantics is simulated, I was wondering if the authors (or others) have previously run Kimera-VIO in Gazebo, or other simulated environments. If so, any tips or advice for setting up the simulation and generating suitable artificial data would be greatly appreciated!

In particular, I am currently stuck on a QueueSynchronizer error when I try to run Kimera-VIO "online" as my simulated cameras are not publishing synchronized frames.

F1130 15:19:51.386656 10433 QueueSynchronizer.h:138] Check failed: timestamp == payload_timestamp (7061000000 vs. 7062000000) Syncing queue data_provider_right_frame_queue in module Data Provider failed;

Interestingly, if I save a rosbag of the same simulated camera trajectory, I don't run into this issue, but I do get warnings regarding:

W1130 18:06:35.835642 18957 Mesher.cpp:1380] Missing landmark information to build 3D Mesh.

and

W1130 18:06:35.932917 18955 Tracker.cpp:578] 1-point RANSAC (voting) could not find a solution.

Presumably my simulated environment needs richer texture (image below for reference of what I'm currently working with)? Will having a small (~1ms) time delay between the left and right IR cameras have a noticeable impact on performance for relatively slow trajectories? Does anyone have any suggestions for generating synchronized images and IMU data?

Thanks again for the great library, and thanks in advance to anyone that has any tips!

[moksh-401-511]

hi @shonigmann , have you found any way to connect kimera with Gazebo? kindly let me know