Re: [glob2-devel] understanding pathfinding

[Stéphane Magnenat]

Hi Jannis,

You can think of the gradient as a pre-processing step for A*. Using the 
gradient, and considering that there are no dynamic obstacle, the 
pathfinding is optimal. When areas become crowded, the assumption of no 
obstacle is not true as other globules cannot be considered as transient 
obstacles and the whole pathfinding system collapses, hence the jams.

It is not easy to solve this in general with a low-processing load. That 
said, glob2 is so old that sometimes what was once intractable is now 
possible, and adding to global gradients local gradients that are 
recomputed on jams events could solve the problem. One could then assign 
globule either to the global or to the jam gradients, in such a way that 
the jam get solved. This is interesting.

There is a large and vivid community focused on the questions of 
pathfinding in robotics. You might have a look on google scholar for 
such papers (search for path planning).

Have a nice day,

Stéphane

--
Dr Stéphane Magnenat
http://stephane.magnenat.net