[irteus/irtmodel.l] support :fat option in self-collision-check (#147) #234

k-okada · 2015-07-20T04:15:28Z

snozawa · 2015-07-20T04:52:14Z

Do we need to re-create pqpmodel in changing fat?
https://github.com/euslisp/jskeus/blob/master/irteus/pqp.l#L58
PQP model is updated only when the links does not have pqpmodel.
So, when the link has pqp model and we specify fat, new pqp model is not created and fat seem to be neglected.

k-okada · 2015-07-20T05:03:48Z

This method evaluate (original 'fat' to create pqpmodel + distance returned
pep ) <( fat given in this method)

So if you create pqpmodel with fat=0, it works correctly
If you create pqp mode with fat somewhere, it is not
and I'm afraid we do not have mean to know fat value from pqp model

Fat in pqp model means the mergen to create collision model so this can be
hard safety mergen

Fat in this method is threshold distance to consider self collision, so it
may change depends on situation

Maybe we should change key name from fat to min-distance or something

2015年7月20日月曜日、Shunichi Nozawanotifications@github.comさんは書きました:

Do we need to re-create pqpmodel in changing fat?
https://github.com/euslisp/jskeus/blob/master/irteus/pqp.l#L58
PQP model is updated only when the links does not have pqpmodel.
So, when the link has pqp model and we specify fat, new pqp model is not
created and fat seem to be neglected.

—
Reply to this email directly or view it on GitHub
#234 (comment).

◉ Kei Okada

snozawa · 2015-07-20T05:10:35Z

irteus/irtmodel.l

   (let ((cpairs) (col-count 0))
     (dolist (p pairs)
-       (let ((colp (/= (funcall collision-func (car p) (cdr p)) 0)))
+       (let ((colp


Thanks, I understand the motivation of this PR.
For this purpose, it would be better to another name for fat as you said.,
e.g., fat => min-distance, because fatting collision shape is not theoretically same as changing distance.

snozawa · 2015-07-20T05:14:46Z

How about adding fat to link's plist and checking it?

(defmethod cascaded-coords
  (:make-pqpmodel
   (&key (fat 0) ((:faces fs) (send self :faces)))
   (let ((m (pqpmakemodel))
         vs v1 v2 v3 (id 0))
     (setf (get self :pqpmodel) m)
     (setf (get self :fat) fat) ;; change here
     ....
     m))
  )

(defun pqp-collision-check (model1 model2
                                       &optional (flag PQP_FIRST_CONTACT) &key (fat 0) (fat2 nil))
  (let ((m1 (get model1 :pqpmodel))
        (m2 (get model2 :pqpmodel))
        (f1 (get model1 :fat)) (f2 (get model :fat)) ;; change here
        (r1 (send model1 :worldrot))
        (t1 (send model1 :worldpos))
        (r2 (send model2 :worldrot))
        (t2 (send model2 :worldpos)))
    (if (null fat2) (setq fat2 fat))
    (if (and (null m1) (not (eps= fat f1)))  (setq m1 (send model1 :make-pqpmodel :fat fat))) ;; change here, check fat change
    (if (and (null m2) (not (eps= fat2 f2))) (setq m2 (send model2 :make-pqpmodel :fat fat2)))
    (pqpcollide r1 t1 m1 r2 t2 m2 flag)
    ))

In this way, we can check fat change and when we increase and decrease fat we can obtain pqpmodel for adequate fat value.

k-okada · 2015-07-20T05:21:55Z

having fat to plist may good (pqpmodel-fat ?) but if someone is using fat for calculating distance between two links, that's mistake. Main advantage of using PQP over RAPID is that it is able to calculate distance between two objects.

I interpret the motivation of #147 to specify minimum distance when calculating self collision , but if #147 is "really" want to change fat, I'd like to know the detail.

snozawa · 2015-07-20T07:53:48Z

Main advantage of using PQP over RAPID is that it is able to calculate distance between two objects.

I see.
As you said, if we use PQP, using distance is quite useful.

@garaemon,
in #147 , do you really want to change fat? or, is it enough to check distance for you?

k-okada · 2015-07-20T11:38:27Z

add fat to pslot -> #234 (comment)

garaemon · 2015-07-20T11:41:17Z

@garaemon,
in #147 , do you really want to change fat? or, is it enough to check distance for you?

hmm, yes distance may be enough for me.

garaemon · 2015-07-20T11:41:56Z

However, if checking collision or not is much faster than computing distance, :fat parameter would be important.

garaemon · 2015-07-20T11:42:25Z

And also, I'd like to manage several pqp models with different :fat parameters.

k-okada · 2015-07-20T11:51:14Z

why?

◉ Kei Okada

On Mon, Jul 20, 2015 at 8:42 PM, Ryohei Ueda notifications@github.com
wrote:

And also, I'd like to manage several pqp models with different :fat
parameters.

—
Reply to this email directly or view it on GitHub
#234 (comment).

garaemon · 2015-07-20T11:52:31Z

for example, in IK just use :fat 0 and use larger value in order to check
collision with environment.

2015年7月20日月曜日、Kei Okadanotifications@github.comさんは書きました:

why?

◉ Kei Okada

On Mon, Jul 20, 2015 at 8:42 PM, Ryohei Ueda <notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');>
wrote:

And also, I'd like to manage several pqp models with different :fat
parameters.

—
Reply to this email directly or view it on GitHub
#234 (comment).

—
Reply to this email directly or view it on GitHub
#234 (comment).

✉︎ from iPhone

k-okada · 2015-07-20T12:04:16Z

In that case we should check 'distance' between a robot and environment.

◉ Kei Okada

2015/07/20 20:52、Ryohei Ueda notifications@github.com のメッセージ:

for example, in IK just use :fat 0 and use larger value in order to check
collision with environment.

2015年7月20日月曜日、Kei Okadanotifications@github.comさんは書きました:

why?

◉ Kei Okada

On Mon, Jul 20, 2015 at 8:42 PM, Ryohei Ueda <notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');>
wrote:

And also, I'd like to manage several pqp models with different :fat
parameters.

—
Reply to this email directly or view it on GitHub
#234 (comment).

—
Reply to this email directly or view it on GitHub
#234 (comment).

✉︎ from iPhone
—
Reply to this email directly or view it on GitHub.

garaemon · 2015-07-20T12:26:48Z

Yes but Collision check might be easier than distance calculation. But I
don't know implementation of pqp

2015年7月20日月曜日、Kei Okadanotifications@github.comさんは書きました:

In that case we should check 'distance' between a robot and environment.

◉ Kei Okada

2015/07/20 20:52、Ryohei Ueda <notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');> のメッセージ:

for example, in IK just use :fat 0 and use larger value in order to check
collision with environment.

2015年7月20日月曜日、Kei Okada<notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');>さんは書きました:

why?

◉ Kei Okada

On Mon, Jul 20, 2015 at 8:42 PM, Ryohei Ueda <notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');
<javascript:_e(%7B%7D,'cvml','notifications@github.com
javascript:_e(%7B%7D,'cvml','notifications@github.com');');>>
wrote:

And also, I'd like to manage several pqp models with different :fat
parameters.

—
Reply to this email directly or view it on GitHub
#234 (comment).

—
Reply to this email directly or view it on GitHub
#234 (comment).

✉︎ from iPhone
—
Reply to this email directly or view it on GitHub.

—
Reply to this email directly or view it on GitHub
#234 (comment).

✉︎ from iPhone

k-okada · 2015-07-20T13:04:13Z

Ok, I created test code to compare collision detection vs collision distance at 95454fd

the results shows there are no difference between collision detection / distance calculation

create model (base line time)   0.074 ((t 0.0 #f(-50.0 50.0 50.0) (0.0 0.0 0.0)))
pqp collision check   0.274 (0.002004 msec/check)
pqp collision distance   0.263 (0.001887 msec/check)
create model (base line time)   0.074 ((nil 50.0 #f(-150.0 50.0 50.0) (0.0 0.0 0.0)))
pqp collision check   0.234 (0.001597 msec/check)
pqp collision distance   0.234 (0.001599 msec/check)
create model (base line time)   0.068 ((t 0.0 #f(-50.0 50.0 50.0) (0.3 0.3 0.3)))
pqp collision check   0.232 (0.001647 msec/check)
pqp collision distance   0.244 (0.001759 msec/check)
create model (base line time)   0.076 ((nil 30.8431 #f(-150.0 50.0 50.0) (0.3 0.3 0.3)))
pqp collision check   0.232 (0.001563 msec/check)
pqp collision distance   0.228 (0.001524 msec/check)

garaemon · 2015-07-20T13:06:28Z

the results shows there are no difference between collision detection / distance calculation

I see, it means that... we do not need :fat option?

k-okada · 2015-07-20T13:15:05Z

oh, you mean add :fat option besides :min-distance at :self-collision-check
method?

◉ Kei Okada

On Mon, Jul 20, 2015 at 10:06 PM, Ryohei Ueda notifications@github.com
wrote:

the results shows there are no difference between collision detection /
distance calculation

I see, it means that... we do not need :fat option?

—
Reply to this email directly or view it on GitHub
#234 (comment).

garaemon · 2015-07-20T13:23:10Z

oh, you mean add :fat option besides :min-distance at :self-collision-check
method?

Yes,

However if we want to use different :fat parameters for the objects, :fat parameter makes sense.
For example, larger values for end effectors and smaller values for torso links.

…isp#147)

k-okada · 2015-07-20T14:29:20Z

@snozawa , please check k-okada@a762650, which i changed from #234 (comment)

@garaemon, hum if you want to do that we can not supply :fat keyward to :self-collision-check as k-okada@4a6c4ee, may be we can use :fat (list 0 0 30 0 40 0 ...) (too complex...), simplest way is to set :fat for each object before we call :self-collision-check, and in this case we may better to remove :fat option

snozawa reviewed Jul 20, 2015
View reviewed changes

k-okada force-pushed the 147 branch from 4817782 to bde2fee Compare July 20, 2015 05:18

k-okada added a commit to k-okada/jskeus that referenced this pull request Jul 20, 2015

store fat value to pslot (euslisp#234 (comment))

247f45f

k-okada added 3 commits July 20, 2015 23:16

[irteus/irtmodel.l] support :fat option in self-collision-check (eusl…

4a6c4ee

…isp#147)

[irteus/pqp.l] store fat value to pslot (euslisp#234 (comment))

a762650

[test/geo.l] add test co check pqp collision/distance calculation time

4ab94ab

k-okada force-pushed the 147 branch from 95454fd to 4ab94ab Compare July 20, 2015 14:17

k-okada mentioned this pull request Oct 26, 2016

substringp関数に対するsubstring関数があると嬉しい時があるかもしれない euslisp/EusLisp#203

Closed

k-okada added the enhancement label Apr 29, 2017

k-okada added the PR/WaitForMoreFeedback label Jul 11, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[irteus/irtmodel.l] support :fat option in self-collision-check (#147) #234

[irteus/irtmodel.l] support :fat option in self-collision-check (#147) #234

k-okada commented Jul 20, 2015

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

snozawa Jul 20, 2015

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

garaemon commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

[irteus/irtmodel.l] support :fat option in self-collision-check (#147) #234

Are you sure you want to change the base?

[irteus/irtmodel.l] support :fat option in self-collision-check (#147) #234

Conversation

k-okada commented Jul 20, 2015

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

snozawa Jul 20, 2015

Choose a reason for hiding this comment

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

snozawa commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

garaemon commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015

garaemon commented Jul 20, 2015

k-okada commented Jul 20, 2015