#robotics Logs

Dec 11 2018

#robotics Calendar

12:51 AM mrdata-- is now known as mrdata__
01:30 AM veverak: rue_mohr: I think I know why it selected that path
01:30 AM veverak: more tests needed however
01:30 AM veverak: and it's not player/stage, but ROS :)
01:34 AM veverak: the columns are too tight to actally generate path between tham that is shorter than right turn
01:34 AM veverak: (for the planner at this stage)
01:34 AM rue_mohr: heh
01:36 AM veverak: it needs improvements
01:36 AM veverak: but I can finally see what it does...
01:36 AM veverak: rue_mohr: it creates random positions and tries to connect them together
01:37 AM Jak_o_Shadows: visualisation is so important
01:37 AM veverak: it connects together 8 closest neighbours, where movement is possible propably in direct line
01:38 AM veverak: condition: at each step on line, body is not in collision with environment and legs workareas are in collision with environment
01:39 AM veverak: I think I should improve condition
01:39 AM veverak: and also skip random point where the density in area is too big
01:43 AM veverak: nah
01:43 AM veverak: fuck the density
01:44 AM veverak: skip random pose if some existing one is closer than 'x' to it
01:44 AM veverak: it now uses normal distribution to generate the random points
01:44 AM veverak: where center of that is pose that is in the middle of directy line between start/goal
01:49 AM veverak: idea is that random poses for the planner should be generate at first at center and than gradually increase to the rest of the map
01:49 AM veverak: maybe I should changed the parameters of the distribution based on actual pose count...
02:11 AM rue_mohr: hmm
02:13 AM tumdedum_ is now known as tumdedum
07:45 AM veverak: hahaha
07:45 AM veverak: so
07:45 AM veverak: the function for checkign collission alongside linear path is simple
07:45 AM veverak: check collision at step 0,1,2,3,4,5,6,7....
07:45 AM veverak: now I changed it
07:46 AM veverak: 0,5,10...1,6,11...,2,7,12,...3,8,13....,
07:46 AM veverak: I suppose this could improve average runtime a bit
07:59 AM veverak: lol
07:59 AM veverak: fu me
07:59 AM veverak: how to measure distance between poses? ( pose = position and orientation)
08:00 AM veverak: dist( pose1.position, pose2.position ) + angleDifference( pose1.orientation, pose2.orientation )
08:00 AM veverak: ^^ this does not work well in case you have radians and meters as units
08:00 AM veverak: on 50cm big robot
08:02 AM veverak: pose 1 meter away with same orientation has this distance same as pose with same orientation, but 60 degrees apart
08:04 AM veverak: I need better distance for poses
08:30 AM rue_bed: your position is from the middle of the body, right?
08:34 AM deshipu: from the middle of the virtual vehicle
08:39 AM veverak: rue_bed: yup
01:39 PM patr0clus is now known as s3nd1v0g1us
08:24 PM rue_shop3: only one of the ups' can take a 2 hour outage