Topological Mapping and Navigation Based on Visual SLAM Maps

A new approach for transforming sparse feature-based into three-dimensional topological maps. This representation is tailored for path planning use.
Submitted to ICRA 2018. (Arxiv: arxiv.org/abs/...)

[@mirellamelo]

Nice work! I'm quite confused regarding the sensor used. I mean, the paper says a monocular camera, but here in the video, we can see a stereo one. Are you guys using a single frame of it? Also, are you using IMU to map too? I took a time to understand it, especially because you're using "visual SLAM" already in the title, not visual-inertial SLAM. Could you please clarify it? Thanks in advance.

[@myperspective5091]

I'll take one. Seems cool.
Now give it the same capabilities that the
Lingo Driod research robots had.
Then give it a chat function and a virtual pet function.
I'd pay $2,000 plus for a robot with these functions.

[@韋柏安-p5p]

Can you step by step display the instructions for constructing these maps?I don't quite know what instructions to enter to complete the map in the movie.

[@ankeyao]

Hi I wonder how accurate is it in outdoor environment?

[@sreedivikalingutla2675]

can you please share the document link for path planning (to a openvslam map)

[@Songfugel]

If this doesn't require gps-location, this technology could be very useful in cave related mapping like the rescue operation of the lost Thai youth soccer team, so they could drill the escape route more accurately.

[@dymolification]

Yes, this technology does not depend on any GPS readings, it is solely based on cameras and inertial sensors.

[@azdeatherage]

I'm a cave diver and attempting to build something that would do this for me..... anyone wanna help?

[@zhangjun2020]

@@dymolification Hi, really nice work! May I ask how do you do the localization in the topological map? From my understanding, localization is done within the feature point-based map (generated by SLAM algorithm). May I ask is my understanding correct? Looking forward to your reply. Thanks!

[@dymolification]

@@zhangjun2020 thanks a lot! Yes, the localization system was using a sparse feature map from Visual SLAM. You can check the maplab repo github.com/ethz-asl/maplab to learn more.

[@zhangjun2020]

@@dymolification Hi, Marcin, thank you very much for your reply and the maplab repo! We will check this immediately.

[@claudiarudolph4771]

Did you also publish the software for this?