PDA

View Full Version : [Discussion] LIDAR Round 2!



nagmier
07-22-2009, 02:22 AM
Ok Ladies and Gents I've been through the net and come across a surprisingly small number of DIY or Homebrew LIDAR systems. While I can understand that on a "hobby" level this may not be practical, I would think a "simple" system could be devised for Medium sized Robots. I was really impressed and dissapointed with the previous LIDAR thread here at Trossen, Impressed with the ability to get MASSIVE detail and Dissapointed by that same fact as Adrenalynn has told, It is unfeasible due to the sheer amount of data.

Now let me clarify what I would like to get some input on, I'm looking at trying to come up with a relatively low cost LIDAR\LADAR system for use on meduim sized mobile robots for navigation and object recognition.

There are a few projects I've seen on the net and some seem practicle some don't? I'm looking at trying to possibly combine a couple of projects I've seen. I'm pretty much open to any and all suggestions.

So far here is what I have in mind for the projects "goals"
Must be mobile for a medium sized robot (I'm considering medium sized something in the 18-24" x 10-24" Wheelbase, this is arbitrary at this point, my point is I"m not looking for something micro sized)
Used for Mapping
Used for Navigation
Used for Object Recognition
Something "controllable" via microcontroller (Not to process the data)
Serial Command set - Possibly over xbee to backend processing if needed/bandwidth permits
Webcam - Used as a sensor or more preferably to build a combination map (I think this is into openCV and another topic though)
Others - ??? Am I a lunatic?



I would like it if this were to result in a usable "product" it be open source both hw and software I really dislike "product" there because it sounds like it needs a price tag, I hate it so much I feel the need to write this comment about it ironically, I'm mearly using product to describe possible software and PCB/Ciruit designs as it will most likely result in both if this actually does end up working out.

Another thing I should mention, I know this is not going to be anything in the category of quick, I expect this to take some time to work out all the kinks and possibly even get it headed in the correct direction. That being said I'm going to continue below with some of the projects I have run across during my search on this subject more recently.


http://www.pages.drexel.edu/~twd25/webcam_laser_ranger.html
This Webcam based laser range finder seems to have some promise but I'm not sure how well this would work on a mobile platform, Now I've also considered as in the Prev thread that the horizontal beam may be better than the standard pin light and if I remember correctly the Hardware store variety dont seem to pack a big enough punch or they just weren't intense enough


This next link comes in where the horizontal line lasers may have failed us
http://www.spikenzielabs.com/SpikenzieLabs/Laser_Display.html
Check out about halfway down they start talking about projecting a horizontal line and the second video shows what seems to be a pretty intense beam.

The following "Blog Entries" are from a user at letsmakerobots
http://letsmakerobots.com/node/2651
http://letsmakerobots.com/node/2989
These seem to be the most sound that I have run across. I would rather not go the PIC
route I would replace it with an Arduino to start with, possibly a better micro if needed. But I also believe that alot of the "processing" can be offloaded via a Serial connection to a PC either locally or over XBee if bandwidth would permit to something like openCV or a raw data processor.

Ideally I would like to see this as a combination picking the best parts and creating sometime "user friendly" and accessable.

Now for the all important... does this sound a little too ambitious?
Again I'm really looking to open this up as a community project, I don't have alot to go on at this point but even so I have some parts on order that should be here in the next few days I plan on starting to play around with some of these concepts and see where it goes.


As far as work done, so far not much :D
I have sucessfully located and disected an 80mm unused coolermaster heatsink with very nice detachable fan (It doesn't have a huge clunky shell, its actually very svelt :P) I've also gotten it hooked up to my arduino and read the "tach" pin on a pwm pin now here is the wild thing I was reading 99-100 steady on 3.3v, when I switched to 5v it dropped to 67-70 range any ideas on that?
I don't really think it matters too much since as long as the reading is designed around that voltage and not if should be fine right? Thats it so far.

I have a couple blank protoboards and a breadboard I bought to play around with this, I have a red laser (non-TTL)(need to replace with TTL?), plus the rest of my evil domination kit!

I see here that as 330am approaches I got quite winded plz excuse the rambling! I may edit it up a bit at the office in the morning

Adrenalynn
07-22-2009, 04:51 AM
I wish you luck with this! Object recognition? I don't think even a $30,000 SICK will give you that unless we're talking resolution measured in meters. Object recognition over XBee? That's going to be some tremendously new billion-dollar-market compression - giving it away will be very generous.

>> http://www.pages.drexel.edu/~twd25/webcam_laser_ranger.html Sure. The math is the same as my initial posts on the subject. Point source or even linear parallax - it'll work fine - for very rough object detection. As long as the objects are smooth, non-reflective, non-black, and otherwise non-light-absorbing.

You'll also need to find some very expensive filters that are precisely tuned to your laser light frequency to give you better response. Realistically - plan on using it in dim/darkened rooms, that don't have humans, pets, reflective or absorbing surfaces or objects.

I think, realistically, a Sharp IR range finder or ten is a much better bet. A single point-source laser rangefinder with a webcam is easy. A linear parallax range finder is moderately tough but doable. What you're asking for is the Holy Grail and you're going to need to invent it or drop many tens of thousands of dollars for the real stuff. There's a reason it costs like it does...

That's my take on the subject - others may differ - and since I've already failed at it, they may have a rosier take on it. You may succeed where I failed - that's what invention is all about! I don't think I'd invest too much time looking for others work - just dive into inventing your own. If others had it working, SICK etal would be out of business by now.

nagmier
07-22-2009, 08:33 AM
Ok thats sort of what I expected, object "recognition" to be something quite a bit out there.

But you have basically confirmed my fears this just proabably isn't doable at the "hobby" level.

The easiest way I can describe it as a DIY SICK but it seems SICK is using the cheapest stuff out there... Thanks for the input! I am seriously having doubts about this being even close to feasible at this point. I also noticed that the Sharp IR sensors seem to be the way to go, Thankfully I ordered a Long Range and Short Range version.

Please if anyone else has input I am still very open to experimenting and playing around with this.

Sparkfun has a green laser module available do you think that may provide a better light source?

lnxfergy
07-22-2009, 09:15 AM
This is probably one of the best "home-built" LIDAR's I've seen: http://www.eng.buffalo.edu/ubr/ff03laser.php

They used a green laser and a linear CCD array (1024x1 pixels) to get a decent resolution. It's basically a really scaled up version of a Sharp IR ranger. They had pretty good results in the maze at Trinity, but of course that's max distance of about 8'. You won't be able to copy this design exactly anymore though, because the company that made the linear array went out of business.

-Fergs

P.S. this lidar was built in 2002-2003, so I'd guess that technology, especially video tech, has come a long way since then (in terms of how many megapixels the CCD has)

SK.
07-31-2009, 04:50 PM
Hokuyo is now offering the URG-04LX-UG01 LRF, a low cost version of the URG-04LX which you often see for research applications. It costs about 1100$. Thatīs probably the cheapest real laser scanner you can get at the moment. Still a quite hefty pricetag, but thereīs hope that in the next years scanners will become affordable.

Well, and then thereīs Microsoftīs Project Natal for the XBox360 coming in late 2010. Now if that thing gets hacked, a 60Hz 3D Camera with a few meters range becomes available for < 200$ (and if itīs working as advertised).

nagmier
07-31-2009, 05:50 PM
Basically I see that as a hobby, Laser range finding is almost impossible, I have some experiments planned to take place over the next 6-9months but nothing serious more for me nothing to make it out of the lab at least I doubt it anyways.

Al1970
08-01-2009, 02:02 AM
Hi. I have to go with Adrenalynn on this. I put together a very cheap system just to play around with. I wanted to see if I could build something better than the Sharp IR range finder for about the same price. I used a gameboy camera. It's B&W, I wanted to keep the data low since I was using a PIC and it is very cheap to get from Ebay. I used a red laser. Yes I am sure a IR laser would have worked much better but to be safe I didn't want to go there. I used a red filter from 3D glasses. Here is what I found: I could see the red laser much better with the filter in the sun light but it did not seem to help the camera to "see" better. The camera could not see the red dot after only about 6 feet. When the light hit a glossy white box the camera would see a bigger dot. If the sun light was on the box the camera could no longer see the dot. Compare that with a Sharp IR range finder: The Sharp IR range finder worked when used on the glossy white box in the sun light. The only problem I found with the Sharp IR range finder was that if it only see the edge of an object, it will tell you the wrong distance. So you have to scan a little to the left and right to make sure you are not just seeing the edge of an object.

Al

Cyclop
01-06-2012, 07:48 PM
There is hope! Have you checked out this Hobby Ladar LDR-M10 from MIREMADI (see link below).

http://www.miremadi.com/styled/LDR-M10.html

Apparently there is going to be prototype demonstration video's posted soon and availability to purchase in a few months.

What do you guys think?


Cyclop

Cyclop
01-07-2012, 04:39 PM
Well the video is up, it shows how fast it reacts to objects.

LDR-M10 HOBBY LIDAR VIDEO

http://www.youtube.com/watch?v=nk2xY3usY0k

Cyclop

lnxfergy
01-07-2012, 08:07 PM
There is hope! Have you checked out this Hobby Ladar LDR-M10 from MIREMADI (see link below).

http://www.miremadi.com/styled/LDR-M10.html

What do you guys think?

Honestly, for robotics purposes, spending $399 on an XV-11 vacuum and tearing out the laser to use is probably more useful. The XV-11 laser has better range (4m vs 4ft) and already has a rotation motor giving a 5hz 360 degree view of the scene.

-Fergs

Cyclop
01-08-2012, 12:56 AM
Fergs,

Checked out your Blog, quite impressive. The reason I developed LDR-M10 was to avoid all the processing needed for triangulation solutions such as XV-11. The LDR-M10 is a true time of flight range finder that, due to its wide beam width (+/- 20Deg), does not get fooled by furniture and sound absorbing material. Once placed on a servo, it can be aimed at the area of interest to look for obstacles.

Because it is only 65 grams and draws 900mW, it can easily be placed on a small robotic platform. I am working on this right now and will up load a video in a few months.

For mapping a room, I have to agree with you. But for navigation and object avoidance, specially outside in indirect sun, I believe LDR-M10 TOF approach is superior.

Thank you for your valuable feedback, BY THE WAY THE ITEM IS NOW ON SALE!

Cyclop

tician
01-09-2012, 10:29 PM
So it has the basic functionality (wide field of view) of a ~$30 sonar sensor with higher accuracy and more reliable surface detection?

The XV-11's LDS produces a 360 degree radial depth map (encoder position in degrees and radial distance in mm), no external processing/triangulation required.

For narrower field of view, there are the ever popular Sharp IR distance sensors ($10-$15) (triangulating calulations done onboard, but non-linear distance to voltage relationship).
And now Parallax (http://www.robotshop.com/parallax-15-122cm-laser-rangefinder-2.html) makes a ready to use triangulating laser rangefinder with UART interface ($130 for 1Hz ranging over 6"-48" and object detection up to 8 ft. Can grab full images from the camera and everything (BOM, Schematic, Gerber files, Source code) is released under Creative Commons Attribution 3.0, so you can even build your own.).

Honestly, it is really cool to see a TOF sensor, but it doesn't seem too advantageous other than its high update rate.

Cyclop
01-10-2012, 09:59 AM
There is a lot to be said for fast update rate. For example the speed of a vehicle is limited by the sensors update rate. Also, for control loop applications, the speed makes the difference between being able to control something or not. Remember that this sensor can work out-doors too. There are specific applications of this technology that will become more clear in the coming months.

One of the reasons I made this sensor was the the limitations of ultrasonic sensors. From a slow response all the way to undefined lobes and limitations on performance on different materials.

Do you know the power requirements of the XV-11?

Thanks for the great feedback.

Cyclop

Laser Developer
01-23-2012, 04:52 AM
it is really cool to see a TOF sensor

Is this what you were looking for?

TOF LRF home made, open source:

http://forums.trossenrobotics.com/attachment.php?attachmentid=3730&d=1327315692


max range: >100m
best resolution: <1cm
best update rate: 100 readings per second