Particle Swarm Optimization with 5, 50, 500, 5000 particles

This is a variation on the Particle swarm optimization algorithm and visualized with OpenGL.

this PSO (and true PSO) is a fairly straightforward algorithm
step 1: initialize all particles to random velocities and accelerations
step 2: determine which particle is closest to the goal
step 3: adjust all accelerations toward the goal
step 4: update particles positions based on velocity, update velocity based on acceleration
step 5: go to step 2.

Traditional PSO adds a bit more complexities such as keep track of your global best position and move to a location that is somewhere between your global best and the swarms' current best. There are a lot of interesting variations that could be performed for this algorithm.

This version of the algorithm also has a random factor into it, during step 3 there is a 70% chance that the particle will not adjust his acceleration towards the current leader. This gives a nice fan out effect as the particles are allowed to carry on with their current trajectory accept for the 30% where they are trying to get to their leader.

This was programmed in haskell.

Category:

Science & Technology

Tags:

• Particle
• swarm
• optimization
• PSO
• intelligence
• AI
• opengl
• glut
• haskell

[robottcli]

Good job. I am wondering how did you display this? i mean what software is used to show the process, though i know a little about how this particles work but i don't know how you made this display. thx

[irnmn42]

@robottcli at the time i was fascinated with the programming language haskell and used an opengl api known as GLUT to simplify everything. So basically to visualize this it was a bunch of immediate mode calls to draw some squares at positions being updated by the PSO algorithm. I then added the obstacles as I just wanted to see what happened when the goal was allowed to be placed around them in this simple 2D system.

[particleswarm]

I can see the PSO problem..at 3.25min. Its movement is linearizing. What could be the causes?

[irnmn42]

@particleswarm i have a max velocity and max acceleration for each particle. It's a very rudimentary physics system honestly as i was just wanting to see what this would look like. But also my PSO does not utilize a global best as the target is allowed to move away from the swarm. This PSO only uses the swarms current best as a guide. Also there is a random element as to whether the non leaders of the swarm will follow the leader or go about their own business.

[berlinlilaz61]

The topic of my thesis is''ontology partitioning shwarm with Paricle optimierung''hast of what you heard?

[irnmn42]

@berlinlilaz61 I do not believe i have heard of this ontology partitioning before.

[MiszTaJo]

did u have the graph of pso?i mean in one dimension only..thanks

[robottcli]

@irnmn42 Thanks so much for your kind and nice reply. Cheers

[irnmn42]

@didingnuriska I have quite a bit of info in the description, what additional information were you curious about?

[didingnuriska]

it's very nice work...

may I have a little information about this work...

thank you for your atttention...