DPRG
DPRG List  



[DPRG] Fwd: Re: LiDAR-Lite 360 degree continuously rotating mount

Subject: [DPRG] Fwd: Re: LiDAR-Lite 360 degree continuously rotating mount
From: David Anderson davida at smu.edu
Date: Mon Feb 16 10:39:19 CST 2015

Sounds good in theory.   I think Paul is toying with the idea of placing 
reflectors at known locations around the yard, so the lawn mower can 
look for those and track it's own location there-from. That might be 
considerably easier than just trying to recognize the environment from 
its measured geometry.

John is also looking at doing the same sort of thing using the acoustic 
reflections of the environment.   Lots of signal processing required, 
but that's what he's good at!

best
dpa



On 02/16/2015 07:40 AM, Kipton Moravec wrote:
>
> Thought I sent this to everybody.
>
> Kip
>
> -------- Forwarded Message --------
> Subject: 	Re: [DPRG] LiDAR-Lite 360 degree continuously rotating mount
> Date: 	Sun, 15 Feb 2015 21:22:39 -0600
> From: 	Kipton Moravec <kip at kdream.com>
> To: 	David Anderson <davida at smu.edu>
>
>
>
> I would think it could be used for localization in the same way wheel 
> rotation is used for localization. It would have the advantage of no 
> wheel slip.
>
> You would map objects as you found them and by tracking them you would 
> determine which ones are stationary and which ones are moving.
>
> The stationary ones will help you keep track of your movement.
>
> Kip
>
> On 02/15/2015 08:26 PM, David Anderson wrote:
>> OK, I see.   Thanks.
>>
>> The intended use is "localiz[ation] based on (lidar) ranges from 
>> objects in a previously mapped space."   So you don't care so much 
>> about objects that you can't detect, but rather localization based on 
>> the ones you can.  Makes sense. Especially if the previous mapping 
>> was done with the same sensor.
>>
>> I hope none of those objects move.  Real world objects (people, dogs, 
>> vehicles, parked cars, other robots) have a tendency to do that.  
>> Seems like that would sort of blow the localization.
>>
>> Of course, you know me.  When you start talking about an outdoor 
>> robot having to run in "previously mapped" spaces, my interest begins 
>> to wane...  ;)
>>
>> Looking forward to your presentation.
>>
>> cheers!
>> dpa
>>
>>
>> On 02/15/2015 08:06 PM, Paul Bouchier wrote:
>>> Hi David,
>>>
>>> All good points, but the intended application is localization, not 
>>> obstacle avoidance (though it may provide a partial solution to 
>>> that. An outdoor mowing robot is most in need of accurate 
>>> localization. Decent obstacle avoidance, particularly of things the 
>>> robot might run over and chew up, are also important. The mower has 
>>> a bumper strip, but that wouldn't stop it getting stuck under 
>>> benches, as John Swindle noted in his excellent roomba/neato/mint 
>>> analysis. I think the mower has roomba-style downward-looking cliff 
>>> detectors. Any decent solution needs multiple defenses, but without 
>>> good localization there's little point in working on the "easier" 
>>> problems because you can't get to a working solution. Tackle the 
>>> show-stopper problems first. The robotics world is missing good and 
>>> inexpensive localization solutions. GPS is a partial solution but it 
>>> will lie to your face and tell you you're somewhere else, with high 
>>> confidence, and make the mower mow down the flower-bed, confident 
>>> that it's on lawn.
>>>
>>> ROS has a particle filter localization module which localizes based 
>>> on (lidar) ranges from objects in a previously mapped space. I think 
>>> a high-speed long-range lidar may have application as a sensor to 
>>> feed that algorithm.
>>>
>>> Regards
>>>
>>> Paul
>>>
>>>
>>> On 02/15/2015 06:38 PM, David Anderson wrote:
>>>> Thanks Paul,
>>>>
>>>> Just trying to get a feel for how this might be used. 16cm at 40m 
>>>> means 4cm, about 1 and 3/4 inch, at 10 meters.  So, unless the 
>>>> scans are jittered vertically, nothing above or below a 1 and 3/4" 
>>>> strip would be detected at 10 meters, which is still 30 feet away 
>>>> from the robot.
>>>>
>>>> That might work well for walls or fences or tree trunks, but would 
>>>> miss many (most?) real-world outdoor obstacles that are higher or 
>>>> lower than a strip a couple of inches wide at 30 feet.   Curbs, for 
>>>> example, or low hanging branches.  The legs of a picnic bench, but 
>>>> not the bench itself.   Chains of a swing set, but not the swing.  
>>>> And so on.
>>>>
>>>> And that little strip then becomes even more narrow as the range 
>>>> closes, i.e., close to the robot, when you need it the most.  At 
>>>> 2.5 meters,  (we're about to collide!) that strip is 1cm, about 1/2 
>>>> inch, and nothing above or below that will be detected.   Not that 
>>>> useful outdoors, in my experience, unless your robot happens to be, 
>>>> you know,  an inch and 3/4 tall.  Otherwise, obstacles that might 
>>>> be detected at 10 meters simply disappear as the robot gets 
>>>> closer.  Not a strategy for success.
>>>>
>>>> So maybe you need to have a fancier rotating mount, one that can 
>>>> pivot vertically to give a denser data set than just a thin strip 
>>>> of the environment at some arbitrary height.   But that would 
>>>> increase the sample time for a complete scan and correspondingly 
>>>> increase the response time of the robot.
>>>>
>>>> If it were operating in a typical artificial indoor robot 
>>>> environment (flat orthogonal walls, "objects" that are simple cubes 
>>>> and cylinders sitting directly on the floor, etc,) it would 
>>>> probably work fine.   Outdoors-real-world, I'd have a lot of questions.
>>>>
>>>> I'm not sure what you can trust from such a tiny, thin-slice sample 
>>>> of the environment.    How are you planning on using this data?
>>>>
>>>> best regards,
>>>> dpa
>>>>
>>>>
>>>>
>>>> On 02/15/2015 03:22 PM, Paul Bouchier wrote:
>>>>> Excellent question David. The Tx spec says 4mrad, so if I'm 
>>>>> operating the calculator correctly, that's a 16cm spot at 40m 
>>>>> (both horizontally and vertically). Are you thinking about narrow 
>>>>> reflectors? Or about the fact that a 1 rps (60rpm) system sampling 
>>>>> at 2kHz would take a sample every 0.18 degrees (3mrad), thereby 
>>>>> giving overlapping coverage?
>>>>>
>>>>> Paul
>>>>>
>>>>> On 02/14/2015 12:39 PM, David Anderson wrote:
>>>>>> Do you what the vertical spread of the beam is?
>>>>>>
>>>>>> thanks,
>>>>>> dpa
>>>>>>
>>>>>> On 02/14/2015 09:46 AM, Paul Bouchier wrote:
>>>>>>> FYI - I saw a post from the LIDAR-lite folks posting on 
>>>>>>> RoboBill's post of a similar rotating platform, and they said 
>>>>>>> that in the first quarter they'll start shipping units with a 
>>>>>>> 2kHz scan rate. Here's the URL - look for the post by Dennis Corey.
>>>>>>> http://diydrones.com/group/ardurover-user-group/forum/topics/contributing-to-giving-sight-to-av-s?commentId=705844%3AComment%3A1870054&groupId=705844%3AGroup%3A903163
>>>>>>> That's more like what we need for a scanning lidar suitable for 
>>>>>>> outdoor work.
>>>>>>> Paul
>>>>>>>
>>>>>>>
>>>>>>> On 02/02/2015 09:45 AM, Paradug wrote:
>>>>>>>> David,
>>>>>>>>       I understand your pain. I often have similar thoughts.
>>>>>>>>      The LiDAR-lite shows promise as a sensor due to its range 
>>>>>>>> outdoors (~40 meters). It only takes a max of 100 points per 
>>>>>>>> second which is slow compared to the rotating LiDARs made for 
>>>>>>>> indoor use, like the Neato XV11 (range ~6 meters indoors), 
>>>>>>>> which make about ~2000 reads per second. In the example shown 
>>>>>>>> in the links, the author is showing a successful 360 degree 
>>>>>>>> sweep at about once every 8 seconds by using 2 sensors. This is 
>>>>>>>> a meaningful step forward for using these sensors usefully in a 
>>>>>>>> robot, but still slow. Using the same idea with faster rotation 
>>>>>>>> and 360 points per revolution, he should be able to get down to 
>>>>>>>> about 2 seconds per revolution.  For a robot going less than 2 
>>>>>>>> or 3 meters per second that might be okay. Especially for the 
>>>>>>>> case where no object was moving (i.e., people).
>>>>>>>>     I am considering an LiDAR-lite based idea that is different 
>>>>>>>> for use on my outdoor robot. The linked project has some ideas 
>>>>>>>> worth stealing ;^). However, I am still on the fence on 
>>>>>>>> purchasing the LiDAR-lite. I need to have a firmer grasp of 
>>>>>>>> what I have to accomplish in the software. The SLAM course I 
>>>>>>>> mentioned earlier is helping that situation.
>>>>>>>>     This might be a good opportunity for a group project within 
>>>>>>>> the club. I know some members have LiDAR-lites.
>>>>>>>> Regards,
>>>>>>>> Doug P.
>>>>>>>> *From:* David Anderson <mailto:davida at smu.edu>
>>>>>>>> *Sent:* Monday, February 02, 2015 9:44 AM
>>>>>>>> *To:* dprglist at dprg.org <mailto:dprglist at dprg.org>
>>>>>>>> *Subject:* Re: [DPRG] LiDAR-Lite 360 degree continuously 
>>>>>>>> rotating mount
>>>>>>>> Cool.
>>>>>>>>
>>>>>>>> <rant>
>>>>>>>>
>>>>>>>> In all my other hobbies, except robotics, when someone has a 
>>>>>>>> new innovation, they show their craft (heli, sailboat, 
>>>>>>>> whatever) doing something really cool, and then explain the 
>>>>>>>> technology that makes it possible.   That's why I might be 
>>>>>>>> interested in the technology, because it allows the platform to 
>>>>>>>> do something interesting that can't be done any other way.
>>>>>>>>
>>>>>>>> But that is never the case with hobby robotics.  Instead, the 
>>>>>>>> technology is demonstrated as if it is an end in itself, with 
>>>>>>>> the hope that someone will eventually do something cool with 
>>>>>>>> it.  Just once, I'd like to go to a link like that posted below 
>>>>>>>> by Doug, and see a video of a robot DOING SOMETHING 
>>>>>>>> interesting, followed by a description of the technology 
>>>>>>>> required to do that interesting thing.
>>>>>>>>
>>>>>>>> (As an aside, this same thing seems to apply in spades to all 
>>>>>>>> the hobbiest ROS robots I've come across.  Surely there are 
>>>>>>>> some HOBBY robots running ROS somewhere out there DOING 
>>>>>>>> SOMETHING  interesting that makes all the pain of getting ROS 
>>>>>>>> to work worthwhile.   If so they seem to be keeping pretty 
>>>>>>>> quiet about  it.   Where are all the videos?)
>>>>>>>>
>>>>>>>> </rant>
>>>>>>>>
>>>>>>>> OK, I'm done, move along, nothing to see here...
>>>>>>>>
>>>>>>>> dpa
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 02/02/2015 08:15 AM, Paradug wrote:
>>>>>>>>> Here is an interesting project for members with 2 LiDAR-lites.
>>>>>>>>> http://www.thingiverse.com/thing:662646
>>>>>>>>> *http://hackaday.io/project/4087-360-degree-lidar-lite-scanner*
>>>>>>>>> Regards,
>>>>>>>>> Doug P.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> _______________________________________________
>>>>>>>>> DPRGlist mailing list
>>>>>>>>> DPRGlist at dprg.org
>>>>>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>>>>>
>>>>>>>> ------------------------------------------------------------------------
>>>>>>>> _______________________________________________
>>>>>>>> DPRGlist mailing list
>>>>>>>> DPRGlist at dprg.org
>>>>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>>>>>
>>>>>>>>
>>>>>>>> _______________________________________________
>>>>>>>> DPRGlist mailing list
>>>>>>>> DPRGlist at dprg.org
>>>>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> DPRGlist mailing list
>>>>>>> DPRGlist at dprg.org
>>>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> DPRGlist mailing list
>>>>>> DPRGlist at dprg.org
>>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>>
>>>>>
>>>>>
>>>>> _______________________________________________
>>>>> DPRGlist mailing list
>>>>> DPRGlist at dprg.org
>>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> DPRGlist mailing list
>>>> DPRGlist at dprg.org
>>>> http://list.dprg.org/mailman/listinfo/dprglist
>>>
>>>
>>>
>>> _______________________________________________
>>> DPRGlist mailing list
>>> DPRGlist at dprg.org
>>> http://list.dprg.org/mailman/listinfo/dprglist
>>
>>
>>
>> _______________________________________________
>> DPRGlist mailing list
>> DPRGlist at dprg.org
>> http://list.dprg.org/mailman/listinfo/dprglist
>
>
>
>
>
> _______________________________________________
> DPRGlist mailing list
> DPRGlist at dprg.org
> http://list.dprg.org/mailman/listinfo/dprglist

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.dprg.org/pipermail/dprglist/attachments/20150216/e01d4f0e/attachment-0001.html 

More information about the DPRG mailing list