#robotics Logs

Mar 07 2019

12:38 AM mrdata||: perhaps they feel impatient or entitled to an answer within 2 minutes
12:43 AM mrdata||: (some dont even wait that long)
12:44 AM rue_mohr: mhm
12:51 AM frosted_flakes: sad. I've been doing an independent study with a robotics professor who is trying to figure out how to make robots more expressive and all of his work is simulation and really abstract stuff because no one has the money or huge amount of expertise necessary to build bipeds which is pretty discouraging
12:56 AM rue_mohr: heh
12:56 AM rue_mohr: I'm controling mine with a human brain
12:57 AM mrdata||: simulation can get you quite far
12:57 AM mrdata||: so long as your model is valid
01:01 AM rue_mohr: ITS TOO LATE THE BRAIN HAS BEEN REMOVED.
01:06 AM frosted_flakes: mrdata|| true. But I feel technology hasn't caught up. High quality sensors for SLAM are super expensive, the torque density of electric motors isn't great, and even preciscion harmonic gear boxes are really expensive
01:07 AM frosted_flakes: I mean hardware implementations aren't super feasible yet and thats mainly what I'm interested in
01:12 AM rue_bed2: openpino
09:31 AM pokmo: hi
09:31 AM pokmo: i read that controlling motor speed by setting power alone is not great for a robot application. anyone know why?
09:33 AM pokmo: the context is PID control
09:53 AM polprog: imagine you put on a blindfold and try to drive your car
09:55 AM pokmo: polprog hmm but motors have hall effect sensors
09:55 AM pokmo: so you can tell how fast the motor's going
09:55 AM pokmo: *turning
09:55 AM polprog: yeah but that is a form of feedback
09:55 AM polprog: setting power alone is no feedback at all
09:55 AM pokmo: if you were blindfolded, you wouldn't have velocity feedback
09:56 AM polprog: you dont know how fast the motor is going
09:56 AM polprog: motors may have those sensors, but you dont magically know how fast its going - you need to read that sensor
09:57 AM pokmo: so usually, what else is needed?
09:57 AM polprog: any form of feedback
09:57 AM pokmo: like an accelerometer?
09:57 AM polprog: https://www.youtube.com/watch?v=UR0hOmjaHp0
09:58 AM polprog: could be - but they arent good at it
09:58 AM polprog: a tachometer is a better example
09:58 AM polprog: this video answers that very question
09:58 AM pokmo: watching now
09:58 AM polprog: and its a series about pid control
09:58 AM pokmo: thanks
09:59 AM pokmo: actually, how do typical self driving cars do it?
10:00 AM polprog: a ton of different systems working in close coordination
10:01 AM pokmo: lidar can't give feedback, right
10:05 AM pokmo: i remember seeing some rovers having just a lidar onboard
10:05 AM pokmo: maybe a GPS too
10:05 AM pokmo: but GPS can only be so accurate
10:10 AM polprog: yeah
10:24 AM goppo: i think lidar-equipped self driving cars determine their position by comparing their lidar mapping with a prior map
10:25 AM goppo: but then i don't know how it's actually done in practice given the constantly changing road conditions, e.g., cars parked by the curb etc
11:29 AM deshipu: the problems is actually that different motors will go different speed with the same power, especially cheap brushed motors will rotate different speed one direction and different the other
11:30 AM deshipu: another problem is that you have slippage on the wheels, especially when turning, and you can't easily measure that
11:31 AM deshipu: various bumps on the road can also make the motor go faster or slower
11:10 PM pokmo: deshipu hmm then how do you get accurate measurements? RPM doesn't account for slippage on the wheels