PC(hemistry)G

PC(hemistry)G is a Unity tool developed to improve the Chemistry TD video game using evolutionary algorithms and other procedural content generation techniques for puzzles. The tool features the implementation of:

1. The Mu + Lambda Evolution Strategy algorithm. This generates routes for enemies to travel along, taking into account different parameters such as map size, route length, and percentage of map coverage, among others. To do this, different routes corresponding to generations are generated. In each generation, the best candidates are selected based on an evaluation with a fitness function. The best-rated routes are part of the new generation, as well as newly random generated maps. The same process is applied iteratively for a certain number of generations, and the best route is obtained according to the given parameters.
2. The Backward from Goal State puzzle technique. This technique was implemented according to the context of the game. At the beginning, the enemies that will appear in the round are selected by the subsequent technique. Each enemy has a certain resistance to the turrets managed by the player. The turrets that cause the most damage to the enemies are determined, that is, the final and ideal state of the game in order to win, or solve it. Then, the compounds that make them up are broken down to obtain the elements with which they are synthesized. Finally, random elements are added to mislead the player. In this way, the initial state of the game is obtained by regressing from the goal state in which the player wins, making the puzzle solvable.
3. The classic local search technique Hill Climbing. This technique was implemented to work in conjunction with the Backward from Goal State technique. It consists of defining the difficulty of the puzzle to be generated based on the parameters provided. The algorithm moves through the space of possible virus combinations, adjusting the amount of resistance and maximum damage that the turrets can produce, so that it depends more on the strategy adopted rather than the characteristics of the turrets.

For more information, check out the project repository on GitHub.