Hands-On Neuroevolution with Python

The maze environment

The maze is defined as an area enclosed by walls from the outside. Inside the maze, multiple internal walls create multiple dead ends with local fitness optima, which makes objective-oriented optimization search not very effective. Furthermore, due to the local fitness optima, objective-based search agents can get stuck inside a particular dead end, halting the evolution process completely. The dead ends are shown in the following diagram:

The local optima areas within the maze

In the preceding diagram, the solving agent's starting position is marked by a filled circle in the bottom-left corner, and the maze exit is marked ...