Increase Flocking Experiment Speed

Requests regarding how to set up experiments in ARGoS.
M_Rizk
Posts: 28
Joined: Thu Apr 30, 2015 2:46 am

Increase Flocking Experiment Speed

Postby M_Rizk » Thu Dec 08, 2016 6:31 am

Hi Carlo,

I am using the code from the flocking example as part of an experiment. In the experiment I have a large arena and I need the swarm to flock to goals generated at random locations throughout that arena. The problem is that the swarm moves very slowly; it takes 2000+ simulation steps to travel approximately 5 squares. Ideally, the swarm should be able to travel to several goals (one after the other) so I can get a good amount of data, and it should do so in a short amount of time so I can run the experiment multiple times. Simply increasing the max wheel speed in the .argos file causes the swarm to fragment. So my question is:

Is it possible to adjust the experiment configuration so the swarm moves faster while maintaining flocking behavior? Alternatively, is it possible to make the experiment itself run faster?

Cheers,
Mostafa

pincy
Site Admin
Posts: 632
Joined: Thu Mar 08, 2012 8:04 pm
Location: Boston, MA
Contact:

Re: Increase Flocking Experiment Speed

Postby pincy » Wed Dec 14, 2016 12:03 am

Hi Mostafa,
Is it possible to adjust the experiment configuration so the swarm moves faster while maintaining flocking behavior?
It's not very trivial to do. Flocking speed is affected by a number of factors:
  • The speed of the robots must be large, but not too large, otherwise the robots risk losing touch with their neighbors;
  • Communication range must be large (to allow for higher speeds), but not too large, otherwise every robot must deal with too many robots around;
  • The sensing frequency affects just how much data you receive per second, which in turn limits the amount of information you can use to correct your direction;
  • Flocking is inherently fiddly, because every robot must juggle between going to a certain place and keeping cohesion. It's not obvious to balance the two, and the right approach depends on the application.
Putting all together, it's relatively normal that flocking speed is slow.
Alternatively, is it possible to make the experiment itself run faster?
Faster than what? I mean, what's the baseline, and what's the configuration you're using? There might be something that we can do.

Cheers,
Carlo
I made ARGoS.


Return to “How to... ?”