Skip to content

INSTALL.md

Latest commit

 

History

History
188 lines (108 loc) · 9.49 KB

INSTALL.md

File metadata and controls

188 lines (108 loc) · 9.49 KB

Baseline: F2F or M2F

Note that:

  • the simplest way to quickly run is that directly download my old session results + new session bag (RB2), performing F2F.
  • harder way is download my old session bag (PK1) + new session bag (RB2), single session mode to build the old session data of PK1. Then use the new session bag to incremental mapping on PK1.
  • The hardest way is to adapt to your own data.

Minimal Case with 2 session data

  1. download data. For demo results you can directly download the old session results.

    old session bag (PK1) new session bag (RB2) old session results
    image (17) image-20250116000038243 image-20250223170438039

You can also use the single session mode to generate the map data of the old session, just set parameters useMultiMode as false. Check the map folder, three files must exists, key_point_frame, pose_graph.g2o and optimized_poses_tum.txt.

image-20250224125924133

  1. set important parameters of file path: save_directory, map_directory,bag_path.

    <param name="save_directory" type="string" value="/home/xchu/data/pose_slam_prior_result/"/>
<param name="map_directory" type="string" value="/home/xchu/data/prior_map/PK01/"/>

<!--set your data bag path-->
<arg name="bag_path" default="/media/xchu/新加卷/HKUSTGZ-DATASET/2023-10-28-19-09-04-Parkinglot-RedBird02/Parkinglot-RedBird-2023-10-28-19-09-04.bag"/>

  2. run launch file, blue trajectoty for the old session and red for the new session.

roslaunch ms_mapping ms.launch

image-20250224012231920

only save data for the new session part, finally use the python scripts to get the merged map for analysis.

rosservice call /save_map

image-20250223172305661

Map Merging With the third session

RB3 data bag PK1-RB2 results PK1-RB2-RB3 results
image-20250122200925827 image-20250327004149392 image-20250327004443119
  1. Now we have merging RB02 based on PK01, we can directly get the necessary data. We have 3760 key-frames with 8820 poses. Note that the final_map_lidar.pcd represents the merged map of PK1 and RB2, but only the RB2 part.

image-20250326233529179

We need manually copy the keyframes of PK01 into the PK1-RB2-TEST/key_point_frame folder, then the number of poses is equal to the keyframes. (you can directly download my data here)

image-20250326234001201

use python scripts multi-session-map-merger2.py to get the merged map of PK1 and RB2. And move it to

PK1-RB2-TEST folder, renamed as map.pcd.

image-20250327001311697

image-20250327001450561

  1. set parameters in ms.launch. map_directory set as the new folder PK1-RB2-TEST.
<param name="map_directory" type="string" value="/home/xchu/data/pose_slam_prior_result/PK1-RB2-TEST/"/>

<arg name="bag_path" default="/media/xchu/新加卷/HKUSTGZ-DATASET/2023-11-26_redbird_03/2023-11-26-19-36-49.bag"/>

<arg name="sequence" default="PK1-RB2-RB3-TEST"/>
  1. get the initial pose. Directly run the ms.launchand quickly stop it to get the first point cloud of PK01 and map.
roslaunch ms_mapping ms.launch

image-20250326235544075

set the initial pose in ms.yaml

image-20250326235800371

run ms.launch again, and this time it will succeed!

roslaunch ms_mapping ms.launch

image-20250327001529771

save results

 rosservice call /save_map

  1. use python scripts multi-session-map-merger2.py to get the merged map again.

    image-20250327003221482

    image-20250327003010034

use scripts multi-session-map-merger_writetum.py to generate the separate trjectory

python3 multi-session-map-merger_writetum.py

image-20250327003352783

use scripts tum-trajectory-plotter.py plot the results:

python3 tum-trajectory-plotter.py

image-20250327003453188

image-20250327003531228 image-20250327003633859

8 Sessions Demo Results

Lets take CP5 as the old session, and use CP2 to do incremental mapping base on it.

  • Build the base map using single-session uncertainty SLAM. It's important to know the cov of each edge in pose graph. Keyframes number must be the same with the poses number.

image-20240806191013216

  • Set the base map as prior map folder, and preparing for the parameters of Ms-Mapping (initial pose). You can use Cloudcompare to align the first point cloud (at the map folder) with the base map to get the initial pose. (This problem can be solved when you integrate a place recognition algorithm into our system).

    image-20240807094852686

    image-20240807094808981

  • Run the Ms-Mapping and save the merged data. The keyframe folder only save the keyframes of the new session data.

image-20240807095156108

image-20240807095347540

  • Use the python scripts to get and visualize all the session trajectory and session map, together with the merged map.

image-20240807101744799

image-20240807101620547

image-20240807101845991 image-20240807110418976

  • set the new base map CP5-CP2 for the next mapping round. You need to add the keyframes of CP5 into this folder. There must be a map.pcd file in the map folder. Check the keyframes number with the pose files.

    image-20240807100001179

Example

We provide example merged data for 8 sessions here, The session order is: CP5-CP2-CS1-CC1-PK1-IA3-IA4-NG. One must clean the separate map to remove the point cloud noise caused by the glass, since this study do not focus on this. The cleaned map also can be download here. Note that these example data may be updated since it is not the best results.

image-20240808085525833

image-20240808085750123

Plot the results:

#trjectory
python3 tum-trajectory-plotter.py 

#map
pcl_viewer merged_map_session_*
image-20240808090003879 image-20240808090223073