The system is formed by self-propelled satellites influenced by the Sun whose objective is not to leave the domain maintaining the maximum possible speed.
All satellites start at the same point with zero velocity, interacting with the Sun according to Newton's law of universal gravitation. Each satellite can auto propel.
Acceleration is calculated by Newton's law of universal gravitation plus the satellite impulse, the velocity is computed from the acceleration and the position from the velocity.
The pulse is calculated by the "brain" of the satellite consisting of a neural network consisting of two hidden layers of 5 and 3 neurons, the input data are the satellite speed, the position of the satellite and the position of the Sun, as output the impulse.
The genetic algorithm changes weight and bias of neural networks by mutation, reproduction and crossing of the neurons, between the top 5 of the satellites. This way, the maximum fitness value is obtained progressively.
Note: The first 5 satellites are not affected by the genetic algorithm. The algorithm randomly chooses two satellites for reproduction and another for crossing, among the first five (they can not be the same).
Fitness must calculate the accumulated velocity (in module) of the satellite "i" in the generation "g", adding all the previous steps "j". Then a constant (0.025) is multiplied at that speed. If the module of the instantaneous velocity is greater than a maximum value established (20), it is penalized by subtracting 10.
It is like example 1 but the satellites interact with three stars.