Diapositive 8 sur 26
But preliminary tests showed this "simplest" PSO is really … too simple. It converges indeed always, but not always to the solution. So I rapidly added two features. The first one is a "no-hope" test. At each time step, we can estimate in advance the maxi
As we can see in the formula, I need the gravity center of the swarm, or, more precisely a gravity center (there are a lot of possibilities). We will see that on a next slide.
Using this kind of test, we waste less time just to go towards a wrong point, and we prepare the next phase, that is to say the second important feature: the re-hope process.