#robotics Logs

Oct 19 2017

12:43 AM rue_mohr: wtf, put a drive in a computer, checked the smart data later, and it LOOKS like it was a brand new 20G drive, 7 power cycles, 47 hours runtime WTF!?
12:53 AM z64555: ?
12:58 AM rue_bed: yea
12:58 AM rue_bed: I just pulled it off the shelf
12:58 AM rue_bed: and its got like _no_ usage
12:58 AM rue_bed: 7 restarts? granted it can take me years to get to 7, but in the process I put thousands of hours on them
12:58 AM rue_bed: its been running in the livingroom for almost 2 days, that means its hour counted had t be damn close to 0 when we put it in
12:58 AM rue_bed: not to mention we power cycled it atleast 4 times
01:27 AM z64555: Well, you are known for buying things, then placing them on a shelf and forgetting about it. :P
01:27 AM z64555: or maybe burying them under another project
03:20 AM Jak_o_Shadows: well, the irlz44n's arrived
03:20 AM Jak_o_Shadows: that was pretty damn quick
03:20 AM Jak_o_Shadows: now I just need the screw terminals to get here
01:51 PM climbing_up: I did some digging and it looks like the Scanse Sweep is using Garmin's LIDAR-Lite sensor
01:52 PM climbing_up: So trying to DIY a spinning platform is not going to be any better
01:56 PM SpeedEvil: what range do you want LIDAR to work to, I forgot
01:56 PM SpeedEvil: and is indoor only OK
01:56 PM climbing_up: 15 meters
01:56 PM climbing_up: Has to work outdoors :)
02:02 PM weyland|yutani: he wants to sense objects size of a car
02:11 PM climbing_up: Yep. Resolution is not as important as range and sampling rate
02:19 PM weyland|yutani: climbing_up, did you find something?
02:25 PM climbing_up: I'm probably just going to order the Sweep and see how it does. I've commented in a few places online asking people for more outdoor range data but haven't heard back yet
02:26 PM climbing_up: This might end up just being a low-range proof of concept since there's no way I can budget for a $5000+ unit right now
02:28 PM climbing_up: I'm bummed there's nothing in the middle price range
02:38 PM weyland|yutani: climbing_up, maybe patent issues
02:42 PM climbing_up: Am I thinking too narrowly? Is there open source software to do low-res outdoor SLAM with stereo vision cameras?
02:43 PM SpeedEvil: I thoguht this was just for a speed camera, or am I thinking of someone else
02:44 PM climbing_up: I assumed it would still be considered SLAM even if the observer is stationary. All motion is relative, right? :)
02:46 PM SpeedEvil: The problem is very, very, very easy if you reduce it to a stationary one involving simple movement gates on a video image.
02:47 PM SpeedEvil: Timing between enters box 1 to enters box 2
02:47 PM SpeedEvil: It only gets remotely complex if you're trying to track highly curving roads.
02:55 PM climbing_up: My goal is to build a speed camera that can track more than one vehicle in an unfamiliar environment
02:55 PM climbing_up: No calibration necessary
02:57 PM climbing_up: It could be a straight roadway, a curved roadway, or an intersection
03:28 PM z64555: sounds like pure image processing
03:30 PM climbing_up: That's what I thought at first, but I wasn't sure how to establish baseline distances without calibrating
03:33 PM veverak: aa:)
03:33 PM SpeedEvil: you drive a known car around, or just average, or ...
03:35 PM climbing_up: Hmm, average what values?
03:36 PM SpeedEvil: Average wheelbase of car fleet.
03:37 PM climbing_up: What confuses me is how to figure out that a vehicle that's further way might have less pixel displacement than a vehicle that's closer, even though the velocities are the same
03:42 PM z64555: which means you have to account for perspective effects (optics)
03:43 PM z64555: You can probably start off with angles, instead of going directly to linear speeds
03:43 PM z64555: and then blend that in once you have a good idea of what the distance to the object is
03:48 PM climbing_up: Damn, I just realized I won't be first any more
03:48 PM climbing_up: https://www.youtube.com/watch?v=mHkz8yT_vxY
03:48 PM climbing_up: ;)
03:48 PM climbing_up: I really want to be able to do this from the ground though
03:51 PM climbing_up: As far as I know, a fully automated ground-based speed measurement isn't available because of the cost
03:52 PM z64555: cost of what, exactly
03:52 PM z64555: materials, or cost of paying the engineers?
03:53 PM climbing_up: Materials
03:54 PM z64555: what exactly are they using?
03:55 PM z64555: lidar, I think you mentioned
03:55 PM z64555: which tbh is the "easy" way out
03:58 PM climbing_up: The pure vision system in the video wouldn't work on a lower-speed environment without overhead mounts
04:05 PM z64555: thats just with single-camera, though
04:06 PM SpeedEvil: It gets more accurate at lower speed as you have more time to integrate the speed
04:09 PM climbing_up: Right. My concern is it's not going to work well with the camera 1.5 meters off the ground
04:10 PM z64555: I'm on a low speed connection, and can't watch the video in a timely manner. So, uh, why would 1.5m above ground make a difference?
04:14 PM climbing_up: I have doubts their "finding vanishing points" and extrapolating road distances based on car lengths/wheelbases will work with such a distorted view of the road
04:17 PM climbing_up: In the video the camera has a clear view of at least 70 meters of road surface
04:20 PM climbing_up: I think I'm ready for the easy way out. This would be by far the most complex project I've ever done :)
06:05 PM i-make-robots: i like fusion360, but i hate editing things from the start of the timeline. "recalculating everything".... every move 3 minute wait...
06:05 PM i-make-robots: https://www.instagram.com/p/BacbfINgWnJ/?taken-by=imakerobots
07:16 PM i-make-robots: when you can't print the part because there isn't enough filament on the roll...
07:27 PM z64555: noooo
07:32 PM i-make-robots: eh, halloween jewelry coming right up...
08:54 PM evilroot: i-make-robots: if you have filament detection you can switch in a different roll
09:16 PM i-make-robots: I don't.
09:16 PM i-make-robots: planning to sell my 2x MK2S and get MK3 in late nov
11:05 PM anonnumberanon: hey z64555 what challenge did you want to see from yesterday, the buckets one right?
11:06 PM z64555: uh, I guess
11:06 PM z64555: you talked about this "really wierd one"
11:06 PM z64555: but didn't go into detail
11:27 PM anonnumberanon: the one with temperatures, freaking hated that one
11:29 PM anonnumberanon: -4 -4 -4 -44 6 3 12
11:30 PM anonnumberanon: determine how many temperatures belong to winter
11:30 PM anonnumberanon: each winter temperature must be smaller than summer temperatures
11:30 PM anonnumberanon: nvm use this one:
11:31 PM anonnumberanon: -4 -4 -4 -44 6 10 12
11:31 PM anonnumberanon: 6 could also be part of the winter according to the question, but you should stop as early as possible, so stop at -44
11:32 PM anonnumberanon: this example should return: 4, for 4 temperatures until summer arrived..
11:32 PM anonnumberanon: z64555,
11:32 PM z64555: that's not a great data set
11:33 PM z64555: especially considering that -44 is pretty friggn cold
11:33 PM z64555: might've been a typo of -4, 4
11:35 PM z64555: on another note, I've determined that beziers are not the greatest thing for path prediction
11:36 PM z64555: quadratic beziers are pretty good for simple trajectory, such as ballistics
11:43 PM z64555: I ran out of paper on my notepad :(
11:43 PM * z64555 grabs another one