Skip to content
This repository has been archived by the owner on Jun 27, 2023. It is now read-only.

Have problem when using Yabloc with own rosbag #63

Open
StepTurtle opened this issue Jun 23, 2023 · 3 comments
Open

Have problem when using Yabloc with own rosbag #63

StepTurtle opened this issue Jun 23, 2023 · 3 comments

Comments

@StepTurtle
Copy link

Hey, great work! It looks very good. I hope it will be implemented Autoware soon.

When I tested the project, I encountered some issues. I followed the instructions in the README and ran the code, but I believe I might be doing something wrong. The algorithm is able to detect lanes, but the trajectory is following the wrong path. You can take a look at the trajectory image below:

Screenshot from 2023-06-23 14-09-36

The trajectory should ideally follow the blue line, but it seems to deviate as if there is a transformation error. I've also included a video demonstrating the problem. If you have any ideas or suggestions regarding this issue, I would greatly appreciate your thoughts.

I just changed image and image_info topics from yabloc.launch.xml and imu topic from twist_estimator_core.cpp. I think yabloc got all of the transformation values from tf_static, I would appreciate it if you correct me if I'm wrong.

Terminal commands I used:

  • ros2 launch yabloc_launch sample.launch.xml
  • ros2 launch yabloc_launch rviz.launch.xml
  • ros2 launch autoware_launch logging_simulator.launch.xml system:=false localization:=false sensing:=false perception:=false control:=false rviz:=true vehicle_model:=golf_vehicle sensor_model:=golf_sensor_model vehicle_id:=golf map_path:=/home/bzeren/autoware_map/yildiz/
  • ros2 bag play bags/yabloc-1/ --clock -r 1.0 --topics /sensing/imu/imu_data /lucid_vision/front_camera/camera_info /lucid_vision/front_camera/image_compressed /vehicle/status/velocity_status /sensing/gnss/pose_with_covariance /sensing/gnss/pose /tf_static
@KYabuuchi
Copy link
Collaborator

@StepTurtle Please try setting the upside_down in this link to false.
In my dataset, the IMU is usually attached upside down, so it is implemented assuming this. I guess the IMU is attached upright in your data, so the rotation estimation is broken.

Originally, I was planning to replace it with Autoware's twist_estimator, so YabLoc's twist_estimator has several flaws. I apologize for the inconvenience.

@KYabuuchi
Copy link
Collaborator

By the way, thank you for your interest in YabLoc. 🙏 I will fix the IMU issue soon.

@StepTurtle
Copy link
Author

I thank you for your great work, and for your response. Making upside_down false solved my problem almost.

Also I am sharing result image with you, if you want to see.

Screenshot from 2023-06-23 15-19-07

There is a little bit error in trajectory, but in my rosbag roads are not good enough. Maybe it cause the localization error, I am not sure.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants