-
Notifications
You must be signed in to change notification settings - Fork 35
/
note.txt
58 lines (45 loc) · 1.91 KB
/
note.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
2023.11.26
kitti2bag 需要用conda 创建python2环境 因为tf只能用于py2环境 一直输入kitti2bag查找缺少哪些包 然后 pip install ** 即可
没有用sc来回环
need a fileplayer
2023.11.27
加入sc回环 ok
2023.11.30
multi-session 可编译
2023.12.2
// TODO:
1)整理 common_lib done
2)icp 位姿变换并发布 done
3)配置yaml done
4)有没有代码冗余,如点云反复变换或拷贝 done
5)能不能保存数据 done
6)设置传感器类型 done
all-api done
2023.12.3
run ok
test ok
2023.12.4 publish msg
loop_marker
trajectory
relo_pointcloud with timestamp
12.10
// TODO:
have we selected sc to get global pose?
how to use teaser ++ to cover poses?
learn imu_lidar_localization?
how to conclude that hdl_localization is poor?
why i can't install ceres?? or glog ?? i have solved it, by refine google-glog + gflags
12.13
global: sc (transform in yaw) + icp to get score, we should publish all candicates !! get a coarse global pose
correct: how to get a precise initial pose?
if the calculated "coarse global pose" is not satisfied, we can select to refine it by manual process. but we wanna use teaser the final pose.
online relocalization: we still use the iESKF in fastlio2, including imu and lidar, but the primary difference is that the measurement function is replaced by the scan2map match between current scan and prior local map.
we guess that the real-time performance is ok, at least the precision of relocalization is perfect like fastlio2.
incremental mapping: we should use the map manager in repo: lidar_imu_localization.
so, the first question is to develop a map manager to manage the prior map.
1.22
there are problems in nearcloud, why it is static??
the mode of lio should be reconstructed
the topics of odometry and path should be refined
1.23
get the cloud in map coordinate from fast-lio2