So I've been working on collaborative localization, which means robots in a team help each other improve their pose estimation!
We submitted a paper to a robotics conference (fingers crossed), and you can find a video from the demo here:
https://youtube.com/watch/...

I watched video and it looks like robot a is mad about robot b, tries to catch him to beat him up, when it fails robot a decides to hit the wall to release anger

@vane Yeah, that's a typical day for a robot.

The wall hit is a good lesson on why open-loop control is bad :)

You do so much cool stuff

@retoor this 😄 genuinely envy your studies/work

@retoor That's just the facade. I don't upload videos of me crying on the floor after 10 hours of experiments end up in the trash because of some logging error.

@NickyBones you should update those videos as well. I'm sure they would get views.

@electrineer 100% but I won't do it for free. Maybe OnlyFans account for emotional breakdown content.

Do the robots make use of SLAM in distributed fashion?

Your bounding Box around the robot is quite big!

@max19931 This is not the real bounding box. It's just an annotation over a frozen frame to help viewers understand.

It's not SLAM, there is a floor plan as can be seen in the visualization.

@NickyBones when is the paper for it coming?

The robot takeover has begun. Make sure you have a tactical nuke in your closet, you'll never know when you need it.

@OmerFlame AI is a bigger problem or better its users and their stupidity.

@max19931 It's already on Arxiv. We would know if it's accepted around July

@OmerFlame No need for nukes, the battery problems are far from solved. Practice fast running for 10 minutes and you are safe.

@NickyBones miniature nuclear reactor in a robot when?

@electrineer Is there a physicist on DevRant we can tag for an answer?

@NickyBones https://arxiv.org/abs/2404.02010, didn't check beforehand. My mistake!

just an idea for effient map merge:

map merge on Lidar range points around the robot by using the radio communication range of the robot as upper bound thereby limiting the search space and alignment testing.

Each robot will treat the others data as how the map data will continue WITHOUT actually having to spend time doing the actual discovery.

Robot send their "full" map state, but the merge will only work on trying to align the map data of the remote robot on data around the remote and only as far as the radio range as the rest can later be aligned with loop closure!

how hard is it to swap the cmos with the lidar(able to run in 360 degree mode than limited as the main boom is covering some of the view) as with will allow faster point collection acround the robot as it done by most others?

speaking of ROS: "slam_gmapping" for local map and "map_merger" to "sync" the local maps across the robots.

How big is the local map on each robot?

would it be easier to just send the latest version of it to the other robot and merge locally?

@max19931 Sending full maps or full LiDAR scans is problematic due to bandwidth issues, which is what we focused on.

There are really nice works about map-merging in multi-robot system that were used for the DARPA challenge. But as a junior researcher, you need to be careful not to go into areas of impossible competition. There are very strong, well-funded groups that dominate multi-robot SLAM and it's really hard to get a paper accepted.

@max19931 There is no local map in our implementation. Only the floor plan as a prior. We aren't building local maps.