Simulating Dynamical Features of Escape Using Grasshopper Physics Engine

Comment

Comment by Nicholas Cecchi on January 10, 2017 at 2:12pm

Saied,

Would you be willing to share your definition? I have a simple simulation working but keep having problems when setting boundaries for collisions. Very inspiring work.

Comment by Ángel Linares on July 28, 2012 at 5:43am

Other possible brute-force and not elegant solution: compute all springs for every moving agent but change the rest-length and/or stiffness of the springs dynamically using vision cone. If the point of a spring in an obstacle falls inside the cone, then the rest-length and/or stiffness rises...

It's not the best solution but it could work (slowly :P). 

Comment by Saied Zarrinmehr on July 28, 2012 at 5:23am

Hi Angel,

I am so glad that I am talking to someone with such an in-depth understanding of geometric modeling! You are right. In fact dividing boundaries to anchor points is not a good choice at all. However, as you have pointed, we cannot dynamically change the anchor-points during the relaxation of the crowd -- i.e. the time span when Kangaroo is running. If we could dynamically change the anchor points, probably we would have considered the closet point of the barriers that is located within the cone of vision as one end of a spring and the moving agent as the other end. This alternative method is also computationally much more efficient since we consider one spring instead of thousands of them. Kangoroo is a great engine, though it has its limitations as well. Maybe designing a battery for the sake of crowd simulation is the best choice.

Please check my second video and share your comments with me!

http://www.youtube.com/watch?v=fPXr0HEwoD8

Appreciate your comments!

Comment by Ángel Linares on July 28, 2012 at 4:48am

You are right...but, you can check if there is direct line (in a mesuareable angle from movement direction vector) between two points and if that is true -> draw a spring that conects both points. The problem is that I suppose that this dinamic list of springs could make Kangaroo crash or don't work properly...I have never try. 

Comment by Saied Zarrinmehr on July 27, 2012 at 2:45pm

Andrew

I agree that I have not provided the sight. However, if there is any difference that would be the "cone of vision", which is set to 2*PI here while in reality it is smaller than PI. Here I did not want to do scripting, but by scripting we can also take hold of the vision angle.

Comment by Andrew Heumann on July 27, 2012 at 12:03pm

That doesn't really have much to do with the way sight functions, since there's no way for it to be obstructed...

Comment by Ángel Linares on July 27, 2012 at 10:15am

Force repulsion I supose...

Comment by Andrew Heumann on July 27, 2012 at 9:06am

How are you simulating the sightedness of agents with kangaroo?