No 0.03 alpha this week as bigger updates are underway: Having enemies design themselves over time using a Genetic Algorithm.
Currently, all enemies in the game are handmade. Crafting each enemy takes a really long time, and it’s time that could be better spent on building other gameplay features.
Besides, seeing the same enemies over and over again makes the Star Squadron repetitive and linear. The Swarm is built from the destroyed remains of ships they’ve encountered, as well as Swarm ships you’ve destroyed, so no two ships should look exactly the same.
The artificial intelligence of The Swarm is also too simple: fly at the player and shoot if the angle is correct. If the player is in the way, keep driving into them.https://youtu.be/
Consider the difference in varying enemy types and artificial intelligence types in Armada on the Dreamcast to the pre-alpha of Star Squadron:
Skip to about 10 minutes in to see some solid gameplay.
Star Squadron pre-alpha footage
There are two choices to help remedy these problems:
1/ Make a lot of enemies by hand, tuning the difficulty manually over a long period of time. This has the distinct advantages of slowing development, having duplicate enemies, and having predictable/dumb AI.
2/ Have the enemies design themselves and learn from their mistakes, building their own ships from bits and pieces of scrap.
Obviously, the second is a much better solution.
How is this possible?
It sounds like science fiction, but it’s a lot closer to science fact.
The plan is to take the 50+ space ship bits currently being used to manually assemble the ships and have the computer build them like a toddler would: mashing pieces together, putting them near the player, and seeing what happens.
Ships which survive longer will have their genetic features carried onto future generations and those that die quickly will be left behind.
Over time, the ships assembled should be quite a challenge and scale perfectly for each player. In fact, because the algorithm will be genetic, the ships featured in the game will be highly unique, lending to replayability.
The aim is to have an early example of this in your hands within the two weeks.