Some projects involving Evolutionary Algorithms and Genetic Programming.
- Implementing a random search algorithm which generates bridges by uniform random placement of coordinates within a predefined rectangular area.
- Implement a basic Evolutionary Algorithm which generates bridges and outperforms the random search algorithm from bridge1a by a signifcant statistcal amount.
- Build upon the Evolutionary Algorithm from bridge1b to create two new Evolutionary Algorithms. The first EA is a constraint satisfaction that considers a new design constraint for allowing boats and other traffic to pass through the bridge. The second new EA is a multi-objective EA (MOEA) that considers both the weight your bridge can support and the total length of material needed to construct the bridge.
- Using the EAs created in bridge1b and bridge1c, implement an island-model EA.
- Implement a random search through valid parse tree space for Pac-Man controllers in GPac, a simplified Pac Man game.
- Evolve a Genetic Programming tree controller for Pac-Man which generates the most high-quality valid action out of the 5 possible actions for Pac-Man that it can find.
- Perform co-evolve within a configurable number of fitness evaluations, a Genetic Programming controller for Pac-Man and a Genetic Programming controller for all Ghosts where a single fitness evaluation is counted as a single full game played between competing controllers.