#robotics Logs

Sep 12 2019

03:06 AM mun: hi
03:11 AM mun: i have a question about SLAM (specifically ORB SLAM). is it common for (ORB) SLAM to be used to just generate a map and freeze the map for path planning? or is the map meant to be constantly updated?
08:31 AM rue_bed: not many of the folks here do that
08:36 AM veverak: mun: that is bad question
08:36 AM veverak: mun: it is used the way is necessary for the task
08:37 AM mun: how come?
08:37 AM veverak: both way (save it || update id continously) are valid and have ceratin benefits
08:37 AM mun: right
08:39 AM veverak: given that I studied orb a bit
08:39 AM veverak: I want to adapt it to my use case
08:39 AM veverak: remove most complex parts of slam (solutions to error being accumulated in the data)
08:39 AM veverak: and just do "local mapping"
08:39 AM veverak: which is continuous
08:39 AM veverak: I suppose that is best for my planner
08:40 AM veverak: (quaderuped legged robot planner)
08:40 AM veverak: it has to be aware of it's surroundings and simplified slam seems as good solution
10:37 AM mun: veverak oh, orb slam actually has a localisation mode
10:41 AM veverak: SLAM = simultaneous localization and ampping
10:41 AM veverak: :)
10:41 AM veverak: *mapping
11:23 AM marder is now known as katsumi
11:38 AM mun: yeah, i mean "localisation only" mode
11:41 AM veverak: ah
11:41 AM veverak: makes sense given how it works
11:49 AM Tom_L: ender better than prusa?
12:15 PM marder is now known as katsumi
06:26 PM logb0t_ is now known as logb0t
07:08 PM logb0t_ is now known as logb0t
08:58 PM lapsang: Hi all. are ultrasonic sensors used regularly for small to medium sized rovers?
09:02 PM zhanx: yep
09:03 PM zhanx: how far out you need to sense can be done with a mix of long and short range and sampling
09:05 PM zhanx: lapsang, with the mix at the same height, you can get a more precise object detection