How to convert GLIM saved map into FAST-LOCALIZATION format (poses mapping problem)

[mahmood-ul-hasan]

Hello, thank you for your great work on GLIM!

I am trying to convert a GLIM saved map into the FAST-LOCALIZATION format, which integrates with HBA. The required format is:

map/
├── pcd/
│   ├── 0.pcd
│   └── 1.pcd
└── pose.json   (each line: tx ty tz qw qx qy qz)

---

GLIM dump format

My GLIM dump is saved in this structure:

.
├── 000000
│   ├── covs_compact.bin
│   ├── data.txt
│   ├── imu_rate.txt
│   └── points_compact.bin
├── 000001
│   ├── covs_compact.bin
│   ├── data.txt
│   ├── imu_rate.txt
│   └── points_compact.bin
├── 000002
│   ├── covs_compact.bin
│   ├── data.txt
│   ├── imu_rate.txt
│   └── points_compact.bin
├── config/...
├── graph.bin
├── graph.txt
├── odom_imu.txt
├── odom_lidar.txt
├── traj_imu.txt
├── traj_lidar.txt
└── values.bin

Each submap has points_compact.bin (point cloud) and data.txt, which includes:

T_world_origin:
  0.894062  -0.445217 -0.0493368 -0.0447592
  0.446668   0.894397  0.0232727  0.0388152
  0.0337653 -0.0428444  0.998511   0.519722
         0          0          0          1

---

My approach

• I wrote a script to:

  • Convert each points_compact.bin into a .pcd file.
  • Extract the pose from data.txt (T_world_origin), convert it to quaternion (qw qx qy qz), and save it into pose.json as tx ty tz qw qx qy qz.

• Example of my generated pose.json:

-0.031940 -0.000772 -0.018251 0.997738 -0.022076 -0.063127 -0.006820
...

• Then I visualized the transformed PCDs using Open3D.

---

Problem

When I open the map in GLIM offline viewer (offline_viewer with optimization enabled), the map looks nice and globally consistent.

But when I visualize my converted format, the map is distorted after applying the poses from data.txt.

---

Questions

1. What is the correct source of ** submap poses** used for the transformation of submap .bin for export?

   • data.txt → contains raw T_world_origin
   • traj_lidar.txt → contains timestamped LiDAR poses (timestamp tx ty tz qx qy qz qw)

2. How can I correctly map submap directories (000000/, 000001/, …) to the poses?

3. Is there already a recommended method in GLIM to export submap poses after global optimization (the same poses that offline_viewer applies internally)?

---

Summary

• My script works, but produces a distorted map.
• GLIM offline viewer shows a correct optimized map.
• I need guidance on how to extract the optimized T_world_submap values for each submap so that my exported format matches what the viewer shows.

Thank you very much!