Genetic AI experiment letting agents learn to jump in 3D. Genes control the sensors direction and weights which influence when the agents decide to jump.
-
Implement simulation environment (agents, boxes, hit-testing) -
Control agents based on genes -
Create new generation based on best fitting genes of previous generations -
Showcase the generation performance, emphasis on evolution (graph) - Add more gene-based factors (angle of sensors, number of sensors, speed)
- Make level generation (boxes) create more difficult scenarios
- Make level generation manually possible
- Robot model by Tomás Laulhé, modifications by Don McCurdy CC0.