http://youtu.be/brZ6E_TcKeo

Post a reply to this message

Le 20/02/2012 13:34, Tail Kinker a écrit :
> http://youtu.be/brZ6E_TcKeo


Interesting. Nice.

Nevertheless, I still have that disbelieving effect due to instant
acceleration/deceleration of mass: Let split that Mecha in 5 parts:

* torso
* right foot
* left foot
* right arm
* left arm

Movement of a foot, as I perceived it in the animation, is either:
 * not-moving, on the ground
 * moving full speed above the ground

There seems to be an enormous acceleration at the transition points,
which seems unlikely as the feet should weight.
Also, they have no forward/backward tilt, whatever the ground. (on the
second part of the movie, this become visible as the front of the right
foot stay above the ground)

The cycle of each foot seems to be the mirror of the other (not sure,
frame per frame analysis is not easy on youtube). 50% on ground, 50%
moving. I would expect 52% on ground, 48% moving, with a stable transfer
of weight during the 2%.
The movement of each foot should be something like a sin (slow start,
slow end, fast in the middle) in the horizontal direction. On the
vertical direction, it might be a fast or slow or constant speed raise,
but the down part should look like a fall: vertical speed is something
in constant acceleration, vertical position is somehow -y² based
(inverted parabola)

The arms are moving in mirror of the foot, so far so good. Yet I would
wonder about their speed & movement: arms serve two purpose on walk
(maybe more):
 * dynamic equilibrium (keep the centre of mass moving in straight line)
 * reserve of kinetic energy (keep the centre of mass moving without
acceleration)

As the torso is keeping a straight, flat & linear movements, only the
movements of the arms are available to compensate the foot. Check the
mass of legs, their kinetic energy as well as their momentum, and adapt
to the movement of arms (from my point of view, the arm are lighter,
hence moving further than the legs)

Would the arm & leg be the same distance from centre of mass, as well as
same height, the robot could be modeled as a cross of diagonal from an
initial square. The intersection (centre of mass) would move along a
line at regular speed. the ends of each diagonal would be the center of
mass of a leg or an arm. Once you get the position of the intersection,
and the position of each foot, you automatically get the needed position
of each arm.

Post a reply to this message

On 20/02/2012 12:34 PM, Tail Kinker wrote:
> http://youtu.be/brZ6E_TcKeo

I like it too.
Time to take it to the next level.

I think doing a walking model in SDL is so very hard and this is good 
but needs a few refinements.



-- 
Regards
     Stephen

Post a reply to this message

On 12-02-20 09:38 AM, Le_Forgeron wrote:
> Le 20/02/2012 13:34, Tail Kinker a écrit :
>> http://youtu.be/brZ6E_TcKeo
>
>
> Interesting. Nice.
>
> Nevertheless, I still have that disbelieving effect due to instant
> acceleration/deceleration of mass: Let split that Mecha in 5 parts:
>
> * torso
> * right foot
> * left foot
> * right arm
> * left arm
>
> Movement of a foot, as I perceived it in the animation, is either:
>   * not-moving, on the ground
>   * moving full speed above the ground
>
> There seems to be an enormous acceleration at the transition points,
> which seems unlikely as the feet should weight.

A very good point.  I'll take a look at fixing that with a bit of extra 
math.

> Also, they have no forward/backward tilt, whatever the ground. (on the
> second part of the movie, this become visible as the front of the right
> foot stay above the ground)

That is true, and an issue of which I was aware.  Still working on a fix 
for that.

> The cycle of each foot seems to be the mirror of the other (not sure,
> frame per frame analysis is not easy on youtube). 50% on ground, 50%
> moving. I would expect 52% on ground, 48% moving, with a stable transfer
> of weight during the 2%.

Another good point, and one that I think that will be fixed when I fix 
the first point.

> The movement of each foot should be something like a sin (slow start,
> slow end, fast in the middle) in the horizontal direction. On the
> vertical direction, it might be a fast or slow or constant speed raise,
> but the down part should look like a fall: vertical speed is something
> in constant acceleration, vertical position is somehow -y² based
> (inverted parabola)
>
> The arms are moving in mirror of the foot, so far so good. Yet I would
> wonder about their speed&  movement: arms serve two purpose on walk
> (maybe more):
>   * dynamic equilibrium (keep the centre of mass moving in straight line)
>   * reserve of kinetic energy (keep the centre of mass moving without
> acceleration)

True, but in a BattleMech, balance is maintained with a massive pair of 
counter-rotating gyros.  The arms move because the operator expects them to.

Thanks for the comments.  I'll look into an improvement.

Post a reply to this message

Here's my first attempt at making the leg motion more realistic.  I took 
a look at how human beings walked, and decided that I'd use vrotate to 
track the foot position:

vrotate (<0, 0, 1>, <-360 * Step, 0, 0>) * <0, 0, StepDistance>;

where Step runs from 0 to 1 over the course of the walking cycle, and 
StepDistance is the distance the foot travels over one step.  An #if was 
used to change out the final vector to add a vertical element when needed.

Unfortunately, it looks pretty much no different from the original, and 
has the added difficulty of making it a nightmare to make the 'Mech walk 
over any terrain.  I took a look at my original again, and realized that 
with all the trig functions I was using, I was essentially already using 
vrotate.

Posting this in swf format, because it works so well with looped video.

Post a reply to this message

Attachments:
Download 'arch.swf.dat' (805 KB)