How did you calculate the Position Errors reported in Table 1 in the paper?

[trqminh]

Hi,
I try to reimplement your work in python. However, I don't know my implementation is adequate or not. I see that you reported in the paper that the errors are as in the table below. I wonder how did you calculate them?

Thank you for your consideration!!

[lijx10]

We have VICON ground truth, which is available in our dataset. Tran Quang Minh <notifications@github.com> 于2021年2月2日周二 下午4:58写道：

…

Hi, I see that you reported in the paper that the errors are as in the table below. I wonder how did you calculate them? Thank you for your consideration!! [image: image] <https://user-images.githubusercontent.com/30286786/106575937-303cd380-656f-11eb-9333-293ce93a8f91.png> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AC5G72PX7DHAAGHAJ5TTXYDS465FFANCNFSM4W6HZUMQ> .

-- 黎嘉信 Jiaxin LI PhD, National University of Singapore B.S., Tsinghua University Mobile: 65-9094 1909 / 86-15201519053 Email: lijx1992@gmail.com / jli@u.nus.edu <jli@nus.edu.sg>

[trqminh]

Yeah, I also used the ground truth given. In fact, as understanding your paper and code, I predict the pose through the procedure below

At each update, I compute the L2 distance between the ground-truth pose and the one predicted and updated by reimplemented Fusion EKF. The L2 distance is calculated as follow:

Then I get the mean result of those L2 distances equals 0.35. My question is Can this number compare with the value 0.16 reported?