Resources, The Final Frontier uses a neural network to evolve agents into collecting resources as fast as possible.
This was my thirds attempt. For the first one I had a players controlling a ship (the Enterprise) and there were enemies (the Borg) running after the player, I did not like the result so I went onto something else. My second attempt was to make an agent that could follow a path and apply the neural network for that agent, once again I did not like the direction I was taking that attempt and completely abandoned it.
Then finally…
For this third attempt I used the ship from the first attempt and made agents out of it. Their job is to collect resources on a particular order. The order is decided as soon as they spawn in, a list is created ranking the resources from closest to farthest and that is the order that they must follow to collect all the resources.
The evolution takes place between rounds, each round ends either when 1000 frames have been reached or one particular agent has collected all of the resources.
The numbers on the side ranks the agents according to how many recourses they have collected.
The JavaScript code can be accessed by clicking on js on the top left corner of the embedded Codepen.
The evolution of the agents really surprised me, on one particular run I saw the first round coming out with no agent collecting any resource, and that kept on going for a number of rounds afterwards, until it actually started collecting resources and evolving to collect resources. After it had mastered collecting all of the resources before time running out, the next step in evolution was to collect every resource as soon as possible. It was incredible to see!